Unlocking Government Data: The Secret to Billions in Savings!

manager for public sector field engineering um can I just get a round of applause for such an amazing event so far I think we got another 48 Hours of content here at data AI Summit um I know everybody's pumped and um we want to make sure we give you guys the best and most perfect content that we have here for public sector um over the next three hours that we have everybody together here um so what to expect here at data AI Summit I mean you're already experiencing some of this already and you'll continue to experience this over the next few days but we have public sector sessions 16 customer sessions 30 public sector speakers and during this next three hours you're going to hear directly from our customers they're going to be talking about modernization Legacy migration Unity catalog data sharing gen security um we have a dedicated public sector networking area on the show floor industry dinners happy hours it's going to be a fantastic time for you to collaborate have deep conversations and learn what's happening in our space um and then from an industry Solutions Hub perspective there's demonstrations whiteboard solution sessions technical experts and and demos around llm rag chatbots uh we have a Navy planned and predictive maintenance demo and also a public health gen demonstration as well too we want to thank our industry sponsors um particularly for this session we want to thank AWS um for being our industry sponsor here and really just to kind of kick things off um I think most folks are familiar with

data bricks but some folks aren't familiar with our presence here in the federal Market um datab bricks Federal has been really a key piece of the public sector since 2018 we have more than 150 customers across federal state and local um all the different branches do civilian um and we also have a very large isv ecosystem that Services the public sector as well too so our end customer agencies being able to take the best and brightest of commercial vendors that are building on top of data bricks as well we're a recognized Leader by industry analysts um everybody from Gartner Forester multiple different categories uh from database Management Systems data science and ml platforms eii Foundation models as well too and those will be talked a little bit further as we go forward with the uh keynote session uh tomorrow and Thursday but then the other aspect too is touching user populations like the dod we have 10,000 customers globally in every DOD service Branch touching a data bricks instantiation so the scale is something that you going to hear today is very very impactful and our sled customers are also leading the way as well across education state government local government uh State of California state of New York really large scale organizations that are leveraging data bricks for Innovation and so a couple key announcements we wanted to talk a little bit about uh so we have a new SAS offering so it's datab bricks on AWS govcloud so that's the datab bricks intelligence platform being able to have that for highly sensitive unclassified workloads and so that's currently in public preview and has an agency ATO today and so that's a key key capability I think we's get a round of applause for that I think it's a big technical um Innovation that we're bringing to the market um from a SAS perspective uh So currently in in public preview atto in April 2024 um features including as our customers want theb SQL Unity catalog fully hosted AWS govcloud US citizens us soil fully itar compliant and then we have several upcoming authorizations and certifications uh feder amp High uh which is going to be announced for GA here coming up in July uh there'll be another session I'll I'll talk about here uh that will go a little bit deeper on our security compliance we got DOD I5 for E2 SAS imminent as well too in that timetable and then we also have Hippa all the different stock and ISO compliances as well so there's actually going to be another session that's happening right now during this session but a repeat of this but for folks you know in our space security is first and foremost it's very very key so Enterprise gr security compliance on the data bricks intelligence platform filipo sarini will be driving this particular session tomorrow from 4 to 4:40 uh so recommend you guys try to catch that if you have any interest further um on on these topics and so a quick overview of the

agenda um we're going to be going through with DOD cdao the chief digital and AI office and so that'll be Alex oul and Cody Ferguson from cdao walking through their Journey walking through the Evana platform and then at 220 we're g to have data as a strategic asset in state and local so that'll be an excellent panel of majja Bill and John oan from state of um California HHS and that'll be moderated by Cheryl miles and then we'll also talk a little bit about Genai and so talking about uh security and so balancing AI outcomes with security governance responsible AI uh Jennifer Storm from US Department of State and Omar Kaja our CSO for data bricks uh we'll be moderating that fireside chat we'll have a quick break at 3:15 and then we'll roll quickly into another panel data modernization in the era of AI and so that will be with Ethan Shen and Joe dioro from uh Department of Veterans Affairs and uh that'll be moderated by Brian Davis and then we'll have USCIS so they'll be discussing their migration from PVC to E2 which is a really big things our customers are experiencing and and really trying to drive capabilities with UC and and all the capabilities of the platform so Sean and puru will walk through that journey and and the outcomes that they've been able to achieve by going through that and then at 4:30 Jude Bole our VP of public sector is going to walk through some public sector Awards so uh looking forward to uh the next three hours and if you give me a a round of applause I want to uh welcome Alex o tool and uh Cody Ferguson from cdao we get to see our slides yeah there go all right uh hey everyone I'm uh Alex OU uh I lead the I am the advana program manager and I lead our infrastructure and platforms division at cdao uh first off thank you all for uh having us speak here and uh my name is Cody Ferguson I lead up our data and AI tools and services uh division within the cdao um this is actually my third time being here third or fourth at this point um always an awesome event hopefully everybody's enjoying it as much as we are um I think we're going to immediately jump into like why the the complexity of the dod um or or how complex it is and why it's uh such a challenge for us to solve at the scale that that we're working at uh and then we're going to jump back to uh the kind of the journey and the evolution of advana and uh where it's going here in the near future um so this slide didn't really it didn't really occur to me how uh how adversarial this might seem um until we were in our uh product uh Advisory Board yesterday and I think there was representation from probably 90% of these companies there um but uh this is a really uh quick way to actually explain just why the the mission of the Department of Defense is uh so complex and so big um as you can see uh the these numbers kind of speak for themselves but um we have huge uh Logistics and supply chain challenges we have huge financial and asset tracking challenges across all of our services uh obviously with covid we saw some of the challenges with uh medical and uh medical equipment there um but this is a great slide to kind of level set um the problems that that we're facing within the Department of Defense yeah so so what we did

essentially to to solve this problem is or or what we did was a create essentially create a platform of aenna to start bringing data together uh we started actually with the the infrastructure and Cloud teams to start building a foundation start putting guardrails start bringing in tools uh for us to integrate and bring in data together uh and then we start adding in the data governance teams uh the cataloging teams quality teams uh to start pushing out quality data to DOD um and and providing some of those scaffolding tools and uh or or tool sets for us to give people a head start so we weren't maintaining or building multiple data science platforms and then we started essentially federating that out to our customers um so uh where we aren't the only ones building anymore it is a lot of our customers uh either building where we still build the Marquee products for the joint uh war fighter but then we have essentially our customers from the community spaces which are our uh essentially many instances of advana uh for Navy or for air force uh we have all the combatant commands all of our uh our business areas Building Products inside of advana and sharing those among each other and breaking down some of those silos um and totaling over essentially uh bringing in over 500 data sets uh creating 300 over 300 product lines and uh and serving over 100,000 users inside of the department right now that was a big milestone for us yeah uh to do that we've brought in a number of tools uh right now we're on the data bricks PVC version uh obviously evaluating E2 and the migration there which uh will be coming ideally uh quickly uh but but laring using it largely for uh ETL processing uh a lot of uh machine learning algorithms being built um uh data Federation data Federation we have a unique challenge in the Department of Defense uh where we also operate across three different classification environments so we have to have cross domain Solutions in place to be able to move data back and forth uh to the necessary classification level as well and so when we take a look now that we kind of explained of where where we started from have we started as a single platform really designed around audit bringing financial data aggregating financial data standard in and preparing for the audit so we moved into where we are now of supporting combatant command starting to move into that self-service and supporting multi-domain operations uh and starting to uh really Empower Federated Builders and moving away from the White Glove uh self-service and then we look towards the future of really beginning a full true self-service model really beginning to build through giops and reusable code of starting to push out the platform to customers and providing the data layer or Enterprise data service uh and so that's what we're going to talk about more today of essentially where we are today some of the values we've seen uh over the last year and then really the future of where we're going wow yeah so uh quite a long timeline here of of our history um I think dating back to some policy back uh from probably before I was born actually but but ultimately uh what what we uh set out to initially solve uh to Alex's point is uh this massive problem that the Department of Defense and other federal agencies uh have a challenge with which is simply uh tying the financial transactions back to uh the line items on your financial statements really uh it's it's actually a requirement for any um any uh public entity um to to be publicly traded um but that requirement didn't really hit the federal government until uh I think in the 90s at some point um and so uh because we have uh well over 30 different accounting systems and uh 50 different feeder systems into the uh into those accounting systems within the Department of Defense that was a huge challenge for us um but it was actually where we showed a lot of our initial success uh in uh developing a semantic layer to sit on top of all of that data and bringing it into uh a centralized data uh Warehouse data Lake um and then uh starting to do some of those reconciliations um and and so uh that has been a that had been a huge success for us at that point in time um that quickly translated uh into great if you can do this for financial data can you tell us where all our people are can you tell us the status of our tanks and airplanes and and where they're at what what the uh current Mission capable rates are for any particular unit um things like that became uh immediate uh requirements uh uh from the deputy secretary of defense and and the chairman and uh and at that point in time I think was uh when we also started to uh support covid tracking which uh we're going to jump into some of the other use cases so I won't go into too much detail uh there just yet we'll we'll save that for the ends um but there was a a quite a journey here um and uh we actually were initially aligned to com troller uh which is our our financial organization within the Department of Defense uh we uh started to support these uh Business Health metrics for the deputy secretary to get away from uh PowerPoint presentations uh of uh the current status of these various uh business areas in the department and uh move towards automated uh dashboards based on a on the underlying Source system data um and uh that was massively successful um in fact uh two years ago uh we were uh realigned with a couple of of other organizations to the chief digital and artificial intelligence office uh to ultimately accelerate some of our efforts um so that's that's really where we are

today yeah and what's what's helped us get here uh or or get to where we are is a couple of our foundational pillars and and pillars that we're maintaining um over the next few years but but Staples around uh making sure that that anything in Advan is government owned government secured that we do have essentially government representation in the loop and that we aren't uh just sending it off to to any one vendor or any one contractor um that we're we're are working with customers and adding in capabilities that they're requesting uh through our our monthly uh iteration meetings with them uh working at more of a scaled agile framework so as new things coming on we can pivot quickly and meet the requirements um as required and then really building that open architecture um and making sure that we are anything we are building is open to the government is open to be shared across the government any data products are are available for Federation and shared within any product uh not just one uh one or two tools and then uh really making sure we're providing that lineage through the like through our foundation of audit and the lineage of making sure we're providing a single source of Truth for any data set for any dashboard being uh developed and built and then uh really making sure that we're adding in tool sets for the war fighter uh so we're not building in uh just AI tools or just coding tools that we're adding in uh full full Suite capability that we're adding in uh Auto a AIML that we're bringing in um tools that are more designed for an analyst and things like that so as we kind of continue going forward a big focus is us is expanding the the uh the capabilities of every every analyst which is also somewhat of a recruiting strategy for us as well if you think about it um you know the folks that actually have the expertise to do these things don't want to work in Legacy technology uh shocker I know um so by bringing in these uh newer like industry um accepted capabilities uh we're also uh able to show the the newer generation uh that a career in in Federal service isn't so bad um we actually get to do uh some fun stuff we were just in Germany last week um and we're here at at data bricks today so um yeah it's uh for us it's uh been massively successful just not not just from like uh an architecture and like how we do business uh change but also from a cultural shift within the department yeah and so two of our customers as we mentioned we had Community spaces working with us and Community spaces deploying and building so two we wanted to especially highlight of of the value that we're seeing of moving outside of us providing White Glove service and starting to Federate that out to the customers uh which we have 10 as of now but two highlighting is Jupiter which is the Navy's uh exanti of ADV uh where they have uh almost uh believe it's uh 10,000 uh developers and customers um working inside the platform d today uh building applications I think there are over 300 products now or U being built in there using over 100 data sources from from Navy that's been ingested and standardizing to even doing close to real-time replication of their Erp system and bringing that into Advanta for a a near realtime uh status of funds and and what they have and then our Air Force Community uh the blade A4 team uh working to really modernize and bring up some of their systems uh developing uh and migrating essentially some of the the architecture systems that we have and bringing them to a more modern and open architecture skip ahead a yeah there we go uh and then there's four uh big use cases we wanted to share bring out uh of of where Advan has been working and and and going over over the last couple of years uh so so the first one uh is our improper pay so started in the audit obviously uh but but really analyzing all of Dodd's transactions uh identifying risky or or or potentially uh improper charges uh flagging those for the analyst over at dest and then being able to recoup those costs so over the last two years uh being able to recall or recoup almost $1 13 billion uh worth of money uh is a is a big Flagship for our financial team if that were taxes we'd actually get a cut of that yeah no that that's a massive I'm I'm actually just excited we can even share share these with you this year I know in the past uh while we were working these day in and day out there was a little bit of sensitivity to actually even sharing them to the the general audience here so uh this is awesome yeah and next uh obvious co uh when everyone shut down uh we did not uh we we kept working uh we had military members that still had to be reporting to duty still had to be moving uh in and out of of their locations uh so uh we had we stood up essentially a covid tasking Workforce bring in all the logistics data all the PPP data all the supply chain data all of our HR and Personnel data uh to start tracking uh uh user our our members and our our service members throughout um throughout our Enterprise and making sure we were advising properly of where we needed to uh Implement additional restrictions or or where we could return to work yeah no that that one didn't prevent us from going into work unfortunately but um it was uh it was a great use case uh for us because it was really uh really pushed us outside of that Financial uh business area um to to move into Logistics and supporting the war fighter um so it was really one of our first uh one of our first use cases where we got outside of our comfort zone a little bit and uh showed some some massive success based on the underlying architecture we had in place yeah and and then the next one was really our first uh one of our first times into real operational uh working with our members uh more closely but really the Afghan Neo of of bringing in uh essentially uh emails and PDFs of uh of uh essentially our partners and um that were requesting assistance uh being able to to help uh advise and and make sure we were connecting uh those uh th those uh Partners to to the right people uh being able to do uh pass information along for Logistics and for Supply so as we started uh moving them out of Afghanistan we were able to make sure the bases had the necessary supplies that we needed uh the covid shots so the ppbe it all tied together but making sure we were able to start evaluating and pre staging supplies uh there was was a big one for our Logistics and and Afghan uh support teams yeah and that one was uh particularly challenging because uh getting out of Afghanistan was uh all military flights um but then uh once those flights landed in in other countries uh we then used contracted flights uh through commercial uh Airlines and so legitimately at one point uh we were watching uh some of our our service members track these flights from flight tracker on an Excel spreadsheet and uh providing the status of when and where they landed uh to the the chairman and uh we're like hold on we we can actually automate this for you guys why are we doing this uh we can tell you exactly when those flights landed put it in a dashboard we can actually track the flights in real time um and and uh and then get the Manifest of who's on those flights uh and bounce those against uh particularly if if those uh folks were coming back State Side boun SLS against terrorist watch lists uh and other things uh to to make sure that uh we weren't introducing any any additional risk uh State Side so um that was a huge challenge for us yeah and the last one um is the Ukrainian crisis of of being able to one have all the logistics data have all the Readiness data being able to see inside of our supply chain uh to advise essentially where we have uh Munitions or where we have capabilities that can be transferred and really being able to aggregate those recommend those um up to to be transferred and then the financial tracking behind it to actually report back to Congress what we are transferring the cost of it to replace uh and and what's needed uh for our partners there uh to continue fighting uh I think is a big one for us yeah and this one in particular I think has uh pushed us even further um into some complexities that we didn't have to deal with in some of these other use cases where we are now working with uh Nations that are not part of NATO uh not formal uh allies or Partners um and so th that introduces some some additional risk as it pertains to the data that we share um but additional challenges in terms of how they access the data that they can see yeah so as as we kind of build off

our experiences and where we go uh with the uh with with what we've learned where the data we brought in is really now beginning that that pivot of advant and changing the way that we do business operations in the traditional uh we we've usually been the we're bringing in the data we're providing the tool set and we're we're essentially building algorithms or we're building dashboards on top of it in a in a very white glove service where we're bringing in contractors and developing or our community spaces are bringing in contractors and developing on the platform uh as we've matured and as we've gotten to a point where we started to hit kind of critical mass it's been a time for us to really examine and and and re-evaluate the platform in its direction uh and and really we've come to the conclusion and and after working with a lot of our partners that building a more open framework uh and building one where we release some of the control out to the the services and out to our comat and commands and partners is really important for us uh so over the last sess a year our team's been working uh very hard at building essentially a software p uh pipeline a container hardening uh pipeline a model hardening pipeline uh a essentially an infrastructure pipeline uh so we can allow essentially development on I 2 uh from essentially any partner uh that then can integrate their tools or their services into the platform and then deploy out through uh essentially I 5 I6 and I 7 from the low side uh using just our core team to to facilitate the deployments On The Higher Side um this is and then we are also working with some of the more open our other Platforms in DOD uh and some outside of DOD to start integrating in so instead of us like forcing or having data migrate to us being able to serve in the data directly so the Investments you've already made or the Departments already made in platforms can continue and we can just serve the the data in a standard way uh with DOD with standard attributes with standard governance models but really started to acel accelerate some of the development and how we're doing it is is we started at essentially the infrastructure level uh we we started with the clouds of building uh essentially abstracting the networks abstracting the clouds and abstracting the domains uh so we built what we call as the platform Factory uh and we've started to to abstract and allow those pipelines and create essentially a standard advana uh core and then building the the Enterprise data services so really concentrating I think we put out a lot on uh open data Frameworks and and data mesh Concepts but really working to create those API gateways create the the governance models to start ingesting and pushing data and federating data out uh to the other systems uh and then taking all of the infrastructure that we showed in the original slide and all the tools and building essentially a common advant on that's a push button deploy so it can be deployed out with all of the governance models with all the security controls uh at will and then where we started to allow customers sandboxes to deploy their web applications their custom uh tools their their cost products uh and any models being able to deploy into the into the uh essentially the factory integrating in building out their tools and then pushing those through the uh environment uh without the assistance of our team but making it much more of a service uh model in a Marketplace so we can expand the offerings in sets inside of the app platform so I think we have uh two slides left um and I don't want to spend all the time on them if there are questions in the room uh because we only have eight minutes so uh these are just some quick facts about uh the amount of data and the data processing we're we're doing uh in the platform and some other key metrics um it's probably a bit old at this point um but a massive amount of data um a ton that's being pushed through a cross domain uh even more actually being uh Federated outside of the platform than is actually coming in at this point uh which is kind of a a myth that we hear quite often in the department um but uh if you want to go to the next slide this uh just kind of drives home the point of where we're uh heading uh on the data and AI front in alignment with this new 1.2 operating model um for our data uh in in particular uh very much the same approach to deploying software uh have have a a pipeline that can actually Federate out the ability to um make data connections bring your data in uh ETL it make it available through an API endpoint uh and and serve it uh uh in a model if if needed or inference it in a model if needed um so that's that's really where we're headed uh with with the data and AI portion of of the tech stack um and with that we have seven minutes so I'm sure you guys have questions got one over here and then we'll go to you next PR slides um you had a stag diagram that talked about Mission stag could you please elaborate what comprises that mission stack please right there yeah yeah so in inside of a essentially what we build are if you look at the mission stack uh that is going to be largely uh essentially our core product so if you look at uh data bricks being part of the engine and driving that uh you look at some of the bi tools of click and Tableau uh but it's also going to include a lot of the governance tools uh of am muta uh of calbra uh of tools that we would deploy kind of more close closer to you uh that we would control um essentially the operations governance of the platform in a common way uh that you would build on top of and integrate your tool sets

into talk about um how how'd you get uh everyone to share their data we can we can all relate in when in government how difficult that is right did you have a command on high that says Thou shalt share or how did that work oh man those were some of the most stressful days of my career um so uh I think about three or four years ago now um the deputy secretary of defense came out with a memo called the creating data Advantage memo um which actually directed uh the entire department to start sharing their data not just with with advana uh although it also identified advana as the authoritative uh data platform for the Department um but uh really what that kind of started was uh when we when we started the business Health metrics um and and measuring like the the actual health of the Department's operations uh all of it was fed off of spreadsheets because it was going into PowerPoints by the time it got in front of the deputy secretary or any other leader uh it was probably 30 days old and then they had a question which went back through like a thousand hours of of Staffing uh and only for that data to come back up and and be stale again uh by the time they answer that question for the day Deputy um so we started with about a list a list of about 200 different data sources that were feeding those uh business decisions um and my task was to get access to every single one of them and uh and if I was having challenges whether it was cultural uh policy or um or resourcing uh or otherwise uh that went on a list that the deputy secretary would review every single uh week and uh and that uh at that point I think um started to remove a lot of blockers for us which was great um but also created a massive workload for me and my team which was fairly small at that point in time um and then uh really I think after we got through maybe like the first 50 people stopped fighting us and they they realized the value of it and uh it really did shift the entire culture of the department I so you have more than 500 data sources across domains one of the slides said how did you solve the semantic problems right because you'll have things that have the same name in different domains but they're different things within one domain you'll have something that's the same thing in the real world but it's defined seven different ways and seven different data sets yeah so I assume you have some kind of semantic layer or incredibly good data catalog with like what do you use and how did you make it um to be quite honest we're we still aren't as successful as we'd like to be in that area and I I meant to say this up front um we have 550 data sources now but there's actually over 4,000 Business Systems in the Department of Defense uh and that's not even including like our weapons platform so um we've only cracked a small portion of the overall uh Department um but the semantic layer I mean the first example I have is uh what we did for audit right um because we were getting data from uh Oracle databases Cobalt mainframes sap uh data sources all all in different languages sap is actually in German we found that out through the process um and we had to map those to that semantic layer uh which was specific to uh financial data at that point in time and the the data that was necessary to actually do the reconciliations to the financial statements uh so we started building the the business logic uh on top of every single one of those systems to translate the data into a format that actually allowed us to aggregate all 50 to 80 of those systems into a single place um but we haven't done that for uh for all of our Logistics just yet we we've started all of our Logistics and supply chain systems we just kind of started an effort on that uh we haven't done it for all of our Personnel systems um uh that it's an incredibly uh hard problem to solve hi my name is Emily I'm with ACI um question followup from the data sources so if we were to bring in commercial data sources what would that D how would that directive work with it because the commercial data sources from what we're finding are all saying well if we're all going to connect into advana then how is that going to work if we're selling these 15 different people now but you're going to connect in yeah um I I don't want to like dance around it I mean the the government and the dod in particular buys a lot of the same data sources over and over and over again and we don't even realize it um because Air Force isn't talking to Army uh for instance and uh that is one of the challenges that uh Advan was set out to solve on the infrastructure side so uh it's it's obviously something that we want to try to solve uh on the commercially available data side um in in that environment but um we haven't just yet I mean we we bring in a ton of commercially available data and obviously it's vendor to vendor it varies on uh how broadly we can actually share that data I think this will be the last one per per time uh this is really awesome quos to all of you uh couple questions really fast really fast answers number one do you consider data Fabric and how will you compare it uh compare it with uh uh data meshes and second are you considering jni for the self-service part yeah uh data Fabric and data mesh I I think like they're uh they're they're incredibly um uh important uh the Air Force has a data fabric which is which consists of like six different platforms I believe six or seven at this point um I view them as complimentary to one another really um the data fabric uh really just being a a a bunch of systems that make up a portion of that data mesh um but I don't know I think everybody's definition changes on on those two particular items and I think they're incredibly useful terms to actually move the Department of Defense into where it needs to go in terms of data sharing and Federation um but I'm not necessarily tied to them as uh as Concepts yeah and as far as the uh llms and and gener of AI uh like so we do have uh obviously pipelines now to bring in foundational models bring them in and sandbox them off inside of the environment and we have started allowing essentially the self-service users to experiment uh to learn and to to train and find tune these for uh certain use cases um I think we're still a little ways away from releasing these into production and allowing these to go until we can really verify the accuracy uh and and traceability inside these models I think we're going to be a little bit away but but we are certainly open to uh letting users experiment letting our customers start learning by doing uh just holding off from making decisions off of them thanks [Applause] [Applause] everyone thanks appreciate this

so for our next session I'd like to welcome mha uh Bill K uh John Ohanian and our uh distinguished uh moderator Cheryl miles from data bricks and it's going to be a focus on data as a strategic asset in state local [Applause] government thank you I can't here they just gave us and then what aren't I was a little bit gonna try to bail on these chairs are you up there I think if my goal is not to fall off the chair I'll just say that be careful I'm good thank you well everybody it's really great to see you here today I'm uh Cheryl miles and I'm really honored to welcome our leader from state and local government here today um they're going to share their Journey towards developing data as a strategic asset and I want to say that you know for those of you who work in government or work with government or work for government change is really hard especially I think in government like government um processes and organizations are by Design not made to be nimble so it takes a lot of energy to to move forward these new initiatives so I'm really pleased to provide these leaders an opportunity to share with you a bit of their Journey so let's start by asking you to introduce yourselves tell us what your role is and what you're and what you've done with data bricks and we're gonna start with John sure good afternoon everyone my name is John ohanyan I'm the chief data officer for the California Health and Human Service Agency and I'm the director for the center for Data Insights and innovation I started uh with the state about four years ago uh I'll give a little of my data journey in in a little bit later but uh joined the the state in May of 2020 which you might remember as the early days of covid uh started with the state moved up to Sacramento and my job uh starting at the state was to be the first Chief data officer define the role for the office for the secretary and then to create a startup within the state and the startup was called the center for Data Insights and Innovation we launched it about three years ago and its goal is two prong number one is how do we like this this prior session which I can't wait to talk to them a little bit afterwards uh is how do we take the 44,000 employees that we have at the California Health and Human Service Agency the $500 billion dollar in in programs that we give out and become more data Centric uh more data Centric when it comes to decision-making especially in times of budget which we're experiencing right now and then uh the primary focus is we've all received Services before and it's very confusing when you speak to one person one day and the person that you speak to the next day doesn't have your information we've all been to doctor's offices where you fill out a form and you wonder why doesn't the next Doctor have it and so we've been challenged with uh attacking that and I'll speak about it a little bit so thank you thank you John and MAA would you like to go next sure my name is MAA Adnan um I am the assistant audited controller and Chief Information officer for the Department of audited controller County of Los Angeles um you know LA County as you know is one of the biggest county in the United States based on their population we kid around 10 million residents with uh a $43 billion budget um the department of auditor controller plays a vital role in providing leadership and expert advice to the 42 County departments that we have on their financial and business uh practices so um and on top of that we also centrally manage the County's Mission critical and complex Erp system which includes financials HR payroll budget procurement um and all the HR Personnel stuff so um me as a Chief Information officer for the Department I uh am uh I'm responsible for not only overseeing the it strategy and the IT operations of the department but also to collaborate with multiple departments for a smooth the operation of the Erp system for the county terrific thank you Macha and Bill I think there's more that she needs to say there's always more we left it for you Bill yes yeah um so Bill kho I am the state CIO of uh the great state of Washington anybody from Washington here all right a few um and I would say that um my data

journey is started way before uh the state of Washington um you know I'll I'll talk a little bit about what we all know that work in government is we've collect a lot of data right we collect a lot of data but how much of that data do we use how much of that data really provides insights into how we are or how we should be improving our services we do a lot of counting of inputs very little outcome based type of analytics where we're looking at the impact of our services that we provide our residents and based on that outcome based analysis making adjustments to make those Services better that's what we should be doing in government we're not doing it and so that's our mission in the state of Washington is using data to better serve the residents of the residents of the state um collecting data and doing cool things with data is great but if we're not doing it to improve our services to our residents then why are we doing it um so that is our focus and um you know we're we're just getting started with with data bricks and building an Enterprise platform and strategy for the state we've got pockets of data bricks in agencies like health and and other um Human Services type agencies but really what we want to do is bring data together just like the presentation we just heard just like John's doing in in California Majin is doing in the LA County bringing that data together and and really doing that outcome based analysis uh aggregating the data looking at our how the impact of our services and then how can we how can we serve our residents uh in a in a better way we have we have issues like homelessness poverty climate change all of that we want to use data to look at all of that and and then make good decisions on how we can improve those areas homelessness is a big issue data and data analytics should be all over that in terms of how can we get how can we improve how can we focus our resources in the right areas so it's a passion of mine and always has been so I will be quiet now thank you Bill I love Bill always has his eye on really what the value why are we all doing this and that's for the citizens so thank you for inspiring us um and while each of your organiz ation and roles are slightly different think that your challenges are similar so including you know the complex organizations that you work in the complex data even sensitive data that you work with so MAA we'll start with you like where are you in your data modernization journey and what are some of the lessons that you've learned along the way um before we before I start on where are we on the journey I would like to mention how we got onto that Journey so just like Bill mentioned uh you know we have a lot of data you know my my uh chief on the uh data uh side always says we are we are rich in data and poor in information so with supporting all these countywide applications we have the entire County's financial data and administrative data but are we able to answer the questions that come from the board or from any other the programs that you know we are working on readily available right so we we don't we can't I mean out of the uh with this whole data that we currently have we are producing close to 2,000 reports and um uh outbounds out of data going to different departments but if the board asks me a specific question for to make a decision uh under a certain program I most of the time do not have the data readily available and I have to look into different data sources to dig that in and to combine it to respond back to the board and I think it it it just it just ties to the same uh challenges that DOD was talking about earlier and just like Bill mentioned I'll give that example as a for homeless now so if the board um ask me okay they would like to know how much the county spend on a on a particular homeless project and how many hours is a County uh employees have currently spent on that project I would have to look into two different data sources to get that information it's going to be financials it's going to be HR now if the board gets a little more um into it and says okay I would also like to know what are the mileage claim requests that come from different departmental users who are actually driving to these homeless sites now I have to look into a few more data sources to get that information there so although we have an integrated Erp system the datas the databases at the end are pretty siloed and they have their own data warehouses and that brought us to uh the entire idea of you know in order if we need to expedite or accelerate our data driven decision making and uh Foster a culture for using data as a strategic asset we need to to integrate and unify that data together so it's easily available for quick accessibility and availability um as well as you know uh manipulation of data whenever we would like to create something out of it so that uh kind of was the purpose of our DNA we call it data and analytics project uh that's where we started off with um again you know just because we have a 20-year-old Legacy reporting system system it's very near and dear to a lot of people you know we have uh created these reports that are pretty complex and and critical for the operations for LA County so they are very particular to make sure that uh we do not have any service gaps so we are approaching this very strategically and tactically we I think year or two in data bricks we started off with phase zero uh we started off and phase zero was very much focused on setting up the foundation so we use uh Azure data bricks uh we used uh data bricks Lakehouse for our data workload strategy we use The Medallion architecture for our data governance strategy um and we started off with phase one uh we started bringing in data and started producing results out of started playing with that data I think our two F Focus was on two B major epics one is um the expenditure and budget and the the other one was a 360 on the vendor monitoring so uh at this point we are you know we rolled out phase one we are strategizing for our following phases on it that's so exciting um for those of you who don't know uh Los Angeles County if it were a state it would be the eighth largest state in the Union so it's larger than most of the states so this accomplishment is terrific MAA thank you it's a teamwork and the Los Angeles County team is sitting right here so congratulations to you all [Music] too too you John can I ask you um I'm sure everybody is interested to learn where are you in your data modernization journey and what are some of your Lessons Learned sure so I guess I would just step in and say when I came on board four years ago while there wasn't a chief data officer we have 14 departments and a some of them had Chief data officers some of them had data officers in particular focused on this journey that we're talking about right how do we bring how do we become more person- centered so in a way I'm standing on their shoulders uh because there was a secretary of Health prior to the current Secretary that he was he was all in on Data he actually brought all the department directors together and said we're going to all sign a data sharing agreement and it was things like that that I walked into and had the benefit of being able to stand with that agreement and hold hold people to it uh I want to also just do quick shout out because our chief data architects in the room David S arria and our our uh acting CIO as well as our deputy director over our Hub SEI kot is here and it's because it's a team we know that right I happen to be up here I'm sharing our story but we're we're really fortunate to have this feeling around our agency that we're serving people and while we operate thousands of programs and if you think about the state doesn't operate a lot of those programs right we take money that comes from the feds or from your taxes and we make investments into counties throughout the the state that that do the actual program and we're not you know we we know that but we know as a whole we need to have feedback loops we need to be better technologically connected with our counties so that vice versa we can see what's happening at the ground level and counties might be able to see how they're doing against maybe their peer counties whether it's larger smaller uh different programs and so I think it's it's important to Stage the fact that I'm a pretty impatient person I came from the startup environment and I've only been at the state for four years and so in some ways I feel like all I keep looking at is where we need to go and what we haven't done but I'm pretty proud of what we've been able to do in terms of that relationship building the trust you know we have we have a hub with data bricks and to get folks we don't need all our departments to get on board with data bricks not that they should should or whatnot but it's not our business in that way but what it is important to us is that the business intelligence we need on our clients and that the realtime Service delivery improves is most important and so what I like doing is saying yeah we have our Hub at the agency and I think it's our job to innovate and incubate better ways to do data governance how do we all get on the all the same data standards so we're all talking about the same data when we're sharing it I was very jealous of dod's presentation because while my boss the Secretary of health and humanit services has that very influential title they still are 14 different departments 14 different entities receiving federal funds and while it doesn't sound really sexy or exciting to talk about it those are the barriers we run into every day me on a phone with a privacy attorney trying to say for the benefit of this person who might experiencing homelessness it would be great to share their data can we find

a way in the rules that we probably written way before we even had interoperability right and so it really does stem federal laws that need to change and so we're spending a lot of time on that so you may look at me and you may not see a data practitioner but what you see is a data evangelist someone that needs to simplify the barriers to the secretary to the governor to other department directors to get them on board with where we need to go because I think a lot of the technical uh Technic technical ways to do it is not the issue right technolog is not our issue um resources at the State of California should not be our issue when we're talking about making decisions of hundreds of billions of dollars we should be able to invest a few million dollars to do our data work better and things of that so uh I would I'll just add two things and I'll stop going forward where IC is I would say middle be safe I don't think we're at the far end Advanced and I see a couple of mods from my team so maybe that's good we're not at our base level but it's kind of one of these things that the more you know the more you realize that you need the more you realize the journey is to go and I think that's where we're at we know that we would like uniform data standards throughout our agency how to actually accomplish that we'll take effort and so the approach that we take is on use cases so it's funny that you mention homelessness obviously you've heard our governor talk about it our secretary talk about it there are so many different programs that touch someone who might be experiencing homelessness there are so many funding streams that they may get a b a bump for this and a bump for that wouldn't it be great if anyone that received dollars that was to support someone experiencing homelessness either pre-post or whatnot they all had to share their data they all had to coordinate so that if you see Joe or you see Joe or I see Joe we all are connected and we can work more efficiently so that's uh that's kind of how we're approaching that work thank you John helps oh absolutely I think um you know I've always said if you want to make an impact work for your government that's how you make an impact on your communities and for those of you who don't know the size and scale of the State of California it's the fifth largest economy in the world so this change is amazing I thank you John thank you for your leadership and Bill I don't know if he said I know we're moving pretty quickly and we want to get through this it's a countdown countown and it just got it faster I got things to say you got things to say and I know one of your key things was improving Citizen Services so did you want to touch on that some more sure sure but I I actually want to touch on the former question a little bit too okay is that okay as a moderator you only have like 45 seconds okay so I just want to say something about um another aspect of what we do in government is we we talk about replacing Legacy systems right a lot we don't talk about modernization and that's that's a shift that we're seeing in the state of Washington is modernization to US has many components the customer experience is one big component of that where we want to start the modernization journey and then data is a huge part of that right so it's not just about replacing old systems but thinking about it as we're doing that what is the customer experience what is the journey Maps those that business process change we need to make what is the data strategy that we need to make what's the what's the uh ocm or the change management that we need to have in our organizations moving to Agile incremental phased implementation all of that we're seeing uh all of our large Enterprise projects and data is has to be a big piece of that but we tend to forget as we're talking about repl repacing old systems we tend to just replace it with a new system we don't we don't improve the customer experience we don't even ask our customers sometimes what is your experience I can tell you in Washington when we ask our customers that it's like why do I have to go to 20 different websites to get the information and the services I need I want to deal with the state I don't want to memorize your or structure your acronyms all that crap just give me one face of government and let me seamlessly access those services and so that's what we're after and it's about modernization it's about da data as a foundation it's about really understanding the experience of our residents in the state of Washington and it's diverse right we're not designing systems for the few we're designing systems for everyone in our state and so it's not just digital experience it's office experience as well and so that's a huge focus on for us right now is using data using data bricks data analytics to really understand the customer experience and then as we design and and modernize our systems um ensuring that we're making good decisions because we have the data to back that up look at that I I mean I left some time too and and we only have six minutes left so we're g to jump straight to a a quick wrap up here um and so I'm going to do this sort of round robin really quickly so majaa

we'll ask you first is Gen AI on the horizon and what are you thinking about from the power of gen well or gen ready or not it's here right so and it's and it's expanding exponentially you know from a public sector perspective we are always uh we known to be very cautious and being audited controller we are overly cautious uh thinking about security privacy and everything should we use it should we not um we are currently working on the basic fundamentals of moving towards gen so you know for the success of any gen implementation uh the quality of data is is integral it's it's really important so we are working on uh you know cleaning up data uh data governance you know building uh all those rules policies um tightening it up as well as data classification which is also very important for any um gen model so despite being overly cautious we have spent a lot of time in uh learning and I would say analyzing and making a decision so we are decided to start a pocc with uh using you know data Break state-of-the-art um dbrx model which is uh retrieval augment augmented generation llm it's a rag model uh that we are excited to start a proof of concept on and what we are doing is we have again a weth of information in unstructured PDFs and these PDFs could be functional documents you know a lot of uh County fiscal policies and and you know all that so we are putting that together in an inventory we're putting it doing it in a in our private Network so we make sure that we're not uh doing the worldwide web on that and very securely we are proceeding in that use case so that's one but I also want to touch on the Lessons Learned just just maybe it would be useful for someone in the room um and that is definitely when you implement your lake house make sure that you are using The Medallion architecture to maximize the uh potential of a lakeh house and and secondly do not forget and I think as technologists we tend to because we get too excited about moving on and starting new technology that we forget about the people aspect of it so I believe everyone kind of touched on it and Bill mentioned that we have we can put a lot of data out there but is everyone getting a value out of it so even with modern data platforms even if we get a lot of uh data accessibility and everything our our user receptive to it are our users trained how to use those data platforms and are they also along with us in that Journey so that's also very important something we have learned uh as we were implementing from our phase zero where we got push back from our users in LA county um to get their Buy in as we moved into phase one so we got them from engagement in Phase zero to excitement in phase one as we just change a little bit on our narrative uh to make sure that they're comfortable with it so yeah it's terrific that's the hardest thing to do is to get started John I'm going to ask you what ji use cases are you most excited about and why sure so uh I would just say Let's do an and so I don't have to repeat because I think there's a lot of similarity in in what we're approaching uh I'll just add that the secretary and the governor both you know the governor put out an executive order uh about a year ago this time maybe a little bit later uh addressing this because I think it's that same feeling right we want to harness the best technology to serve our clients the best and I don't want to say having said that I want to say and we need to protect their privacy and I think that's there's a distinguishment between public sector and private sector it's not about knowing the right things to do it's about the political will and the negotiations that you have to go through to make these things happen even with consumer Advocates who will come and say you know we get you're the chief data officer but are you just going to let our data go get out right like there's just this this paranoia scared justifiably concerned uh marginalized communities that have been impacted in negative ways when they've had to supply data in our in our history and so we're just being really cautious of that where I'm optimistic is uh California as well as other states and other municipalities deal with open data and so this data is already out there we've already released it to the public so there's Mass amounts of open data and so the thought and we've been experimenting with Genie right now right like how do we take our current open data and turn it more into a a talking interface because what I want to do is I want to take leadership that maybe aren't so tech-savvy but need answers to questions and let them show show them what's possible with a proof of concept let them deploy something similar in their own or work on a project together but I think that will start getting folks thinking oh I don't have to go to my data department and get a report run I can actually run this uh run this query myself using real language and I think that's the thing that that excited me I say the second part is on the call if you think about the state of California and all the programs we run there's a lot of call centers there's a lot of people that are talking that's a great uh again another another known way that Genai can really help Focus workers on caring for the client and not necessarily taking notes during a call and things of that nature so terrific thank you and I've had a front row seat I think for all three of your data Journeys and I'm really proud and honored to be there I want to wrap up with Bill and I know bill has a challenge um and looking five years out he's challenging and he has a challenge but looking five years out what do you see as the future of so I I'll just say that um you know yes we have to be cautious in government privacy security but let's not be too cautious we are sitting on the precipice of the most amazing technology that I have ever seen in my career and it is going to transform how we serve our residents and how we work and we shouldn't be afraid of that we should not be afraid of that and if we if we don't start to experiment and move forward and get people comfortable uh with AI I then we're going to lose out and when we lose out in government who suffers the people we serve so let's not be too afraid because if we if we don't start to move forward and learn um we're we're just going to we're going to hurt the people that we serve so I say it is going to be the most transformative technology we have ever seen and um let's be a part of that right let's not have it done to us let's be a part of that thank you I have to say I'm uh so appreciative for your leadership really it takes bold leadership to move these initiatives forward and I'm sure all of you are here in the room because you're doing the same so thank you and I invite you to speak to our panelists at the upcoming reception thank you thanks thank you thank you thank it so next up we have a session uh fireside chat for a post honeymoon so it's going to focus on balancing AI outcomes of security governance and responsible AI so uh Jennifer uh Storm from the US Department of State Center for analytics and then Omar Kaja our VP of uh security and field um Chief Information Security Officer with data bricks will be leading this [Applause] conversation okay um got a um got a uh an excited room here guys excited kind of hey look at this okay so I'm uh I'm really looking forward to this session I I've met uh I've met Jen before but I purposely did not actually ask her to give me any of the answers to the question so I can I can be super curious and we can all learn together from Jen and her experience Jen's been at the uh Department of State for a while I was thinking about this the the last person to speak on stage was Bill and Bill was from the from Washington State and now we're going to go to the state department in Washington DC do you see that uh palindrome sad attempt at a dad joke here um but the uh you know what what you te what was teed up by by majda and and Bill and and John before they're talking about their aspirations what they're hoping to get out of AI and they kind of touched upon the fact that there are some challenges along the way and I used to be a real ceso before I I took on my my current role and so I was one of those pesky people that told the business what they can't do and why they can't do it and uh sometimes scaring them sometimes using more measured risk language and so that's likely something that many of you deal with in your day-to-days as well you have this great idea you want to move it forward you want to leverage modern technology modern ways of uh modern ways of working and then things seem to come to a screeching halt it almost feels like there's A Tale of Two Cities going on in

large organizations organizations that have been around for a while that have to do consequential work that sort of feel committed to their mission which all of you working in the uh government organizations that you're associated with likely can likely can relate to so I'm I'm super excited to have Jen share with us what that Tale of Two Cities sort of looks like in in at the department of of State on the one hand you have this excitement about everything that gen can do what Bill talked about how it's the most exciting technology he's ever worked on in his career I hear people say that over and over again many of you how many of you feel feel similarly about AI okay I think we may have some security people in the room but you know on the other hand the security people the Privacy people the governance functions legal compliance they've got a whole lot of reasons why they're not excited and predominantly this is new to them they don't really understand the world of Da data and they don't really understand the world of AI and they're afraid of the things that it may do that it wasn't uh intending to do and so um Jen you you've had a career in data science you've uh started and scaled nonprofits you've uh been in the Consulting World you've been at Department estate for a while in their Center for for analytics serving the state department for their data needs why a career in uh in data science of all the things you could do absolutely um first I just want to share I'm really excited to be here with everyone um you know one thing you said in your intro um which we'll talk about more as we go on um you know the center for analytics was established at the US Department of State which is the oldest cabinet level agency and so you can imagine um that there's a lot of just historic institution process um that you know often comes up against the new practices of science and somehow even though you've been around for 180 years or 280 years you have no Legacy Tech is this true no leg absolutely Legacy no Legacy Tech somehow you uh you've solved that you solved that problem yeah all cutting gon we're gonna talk about that in the next session and how they solve that absolutely um but from a career in data science you know to some extent um you know I started my education background was in economics and data science um and then when I you know first started started working in Consulting and government I worked in it and I worked in really traditional what we thought of as you know systems integration system systems development um you know cost product deployment um and then a couple of years ago I got the chance to um combine my background in data science and combine my background in you know more what had become more traditional it um and got to start working at the center for analytics um over the the past couple years my predecessors had built up a system which we call data. State that's our data an analytics platform at the department um and so I get to work really at the Nexus of the infrastructure layer the application layer and then kind of that data strategy layer um so a little bit similar to the folks from Adan from what they shared we have kind of a similar platform and a similar structure to um really making uh data a part of the culture and a part of the operations at the department um so for me it was really exciting to finally get to bring some of that College learning um back into my day-to-day career um but then really still continue to work at um you know that that it layer which I have really come to love I love those networking challenges those infrastructure challenges um and I like to see them solve to help you know really address those Mission needs yeah I love the uh I love the enthusiasm I think it was that secretary John Dulles who was a secretary of state back in maybe the 70s he said uh it's okay to have problems it's just not okay to have the same problems you had two years ago and I think that's what you're conveying it's sort of the epitomizing that growth mindset of yes you're going to have problems and you roll through them and you go on to the next thing and the next thing and you know one of the things that we find is that uh maybe maybe I'll ask the group what percentage of AI models actually get rolled into production anyone want to take a guess five where where are the optimists in the in the room 45% 45 okay okay overshot too optimistic uh 10% right 10% of models only make it into production so on the one hand we're super excited we got the TR data we trained the model we built the model we're ready to put it out there and for whatever host of reasons 90% of those models don't actually make it into uh don't actually make it into the uh to production environment what what's going on like what you know is it and one of the reasons is the business looks at it and says this isn't as exciting this isn't serving a need how do you prevent that from happening how do you actually build that closer relationship with the uh with the mission yeah I think it's well it's so challenging in so many layers because it's it's easy and fun and exciting to work in a development environment and that's where you want you know a lot of your practitioners um exploring practicing but to actually take a full system from development you know to production is a significant uh amount of work especially at a federal agency because you're not thinking just about does this thing actually work does IT service customers but then can I meet all of the legal the security the compliance mandates that go around it um and can I do that every single year can I not do it once just to get it authorized but then can I continually go back and get it authorized every single year so um it takes a lot of determination and resilience to just bring a system all the way to production um and I think also going through that process you know you'll you'll find either um what you thought was a really good uh use case or a really good solution uh in in your development environment can't meet the standards in Productions or you know just the requirements are to add odds with each other um so uh I think mostly it's just like really going through that that process um and the ones that do make it I think are the

ones that have brought those different practitioners in early so um we work to build really strong relationships with our security our governance our legal our compliance officers so that their requirements get heard and get baked in early so that by the time you go to production it's not a surprise that there wasn't a bunch of you know 20 or 50 requirements that you weren't cognizant of that you didn't know to build towards um so we take a lot of initiative for the ones that do go to production to make sure that those are built in early we're we're going to come back to that uh so some more I want to hear more about any tips and tricks on how you actually navigate those uh those sort of uh wet blankets um one and I can say that because I do security so I can you guys can't say that about security people but I can say it because you know how this stuff works um the um the what what I'm curious about is one of the things that uh you mentioned Jen is this idea of datadriven diplomacy I you know when I think of diplomacy I think dark rooms I think cigars I think black leather chairs I don't really think like data and tables and you know like but but it sounds like I need to probably get out more what um what is datadriven diplomacy it still might be a little bit of the dark rooms and cigars but more and more um you know we're finding that our um our field practitioners need data to conduct their mission to conduct their operations um for those things that we do domestically here you know State Side to serve the rest of the field to serve the rest of our um diplomatic presence we need data to be able to inform um the decisions so whether it's you know just pure management decisions that the state department has to make to support its Personnel or if it's you know foreign policy decisions um you know our involvement in multilateral organizations all of those things are really starting to require data um a good example recently is um one of our climate leaders uh approached us of uh basically needing better tooling to be able to be participating in negotiations about climate uh the climate crisis climate targets um and needing to be able to basically like dynamically put in inputs and then dynamically see different scenarios that would happen based off of those inputs um that went so far as we would have data scientists at the ready um kind of behind the scenes who are able to support uh those deliberations those negotiations so that eventually when you know official was going to the table um to conduct those negotiations they had a robust set of data to support uh to support those negotiations and they could see the different types of futures or the different types of outcomes that were possible um by using data so that's just one use case um where a data scientist was basically actively supporting and data was actively being brought um and aggregated so that they could have productive conversations about um a topic that at surface you know may seem um very much like a a common uh um you know cigars type conversation but really needed to be supported by a robust set of data and informed by robust data wow as uh I was reading something a few weeks ago and it talked about how there's four different kinds of power and uh the four different kinds of powers one is coercive power where you just have authority and you have a stick and you tell people to do it and they just do it which sometimes the US has had and the state department probably gets to wield that uh often or occasionally and then there is a reciprocal power so Tit for Tat I did something for you you did something for me so you negotiated and then the third is this idea of emotive power you listen to me because I'm charismatic I made you feel a certain way um and then the fourth power is a is rational power so rational power is I'm going to go to logic and I'm going to use data to make decisions I'm I'm curious Jen as you think about the tenure that you've had at Department of State have you seen some of these power sort of increase in frequency and usage and some of these decrease between between the four the coercive power versus reciprocal versus emotive versus rational database power I think you know in all different levels of negotiations whether it's you know just internal or even you know broader external so for instance with Putin what do you guys sorry just kidding it's not allowed to answer that so in in all of these different scenarios you're finding um that the folks who are going into those conversations um again whether it be an internal discussion or something larg um you're really looking to um equip yourself with data they're asking they'll ask a series of questions uh to make sure um you know that they are equipped to have that conversation and I think as many of the you know other um folks can attest sometimes those are questions you can very easily ask answer with data sometimes those are questions that require you to join 15 different data sets um and sign multiple data sharing agreements and you know it is a much larger Endeavor than what you initially set out to do but some sometimes that is what you need to do to be able to answer that that really critical Mission question um so I I do think we are seeing more and more a hunger for data in every type of conversation no matter what type of negotiation structure ultimately gets used yeah um there are basically a hunger and a desire to make sure that you're equipped with you know information across the Spectrum um to be able to have those those good conversations wow love it one of thing

that uh intrigued me is Jen you're your your academic training is not just in data science but it's also in managerial econ economics I can't help but think that that's probably helped you get U much of the success that you that you've had yeah well so it's kind of a business economics degree um and I think as we were chatting before the thing I took most away from economics was actually the game theory portions of Economics um the negotiations the um you know how how you work with other people but from a a macro perspective um I think you know economics is usually not people's favorite yes favorite Topic in high school or college or wherever you first approached it um but ultimately you know it is about the study of our communities and how um people work how communities work um and so there's a natural you know influx of data to be able to answer those types of questions um so uh well I did not go down the track of of being an economist um I am Overjoyed that I still get to you know work with some of those principles and tenants in my day-to-day work yeah I know I I love that because on the one hand we wish that data would rule the world I think there's this um there's this uh adage that said uh uh we we trust in God and all others must bring data is it something like that right and and as much as we wish that was true like how often are decisions made in spite of what the data actually says right do you guys come across that every now and then like we have the sense of if we have the data and we use it then we'll guide ourselves to the best decisions and sometimes in spite of that we end up we end up Mis misusing the data and there's a there's a quote from I think it's Richard oovie who's one of the founders of of the whole this whole idea of advertising and he said um uh some people use data like a drunkard uses a lamp poost for support instead of Illumination right so when you have a lot of data we know we can sometimes use the data and manipulate the data to tell the story that we want and so what Jen is talking about how you bookend that so on the one hand yes you need to understand the data science but on the other side you also need to understand the human science and economics and Game Theory is a lot about how things uh interact and how incentives and and systems uh encourage or discourage people from behaving uh behaving in in in certain ways and and behaving in other ways I'm curious Jen if you were to go back into a time machine talk to the Gen Storm from 10 years years ago a younger version of yourself what would you say especially as you think about these you know as the honeymoon period ends and we get past the uninformed optimism and we get to the informed pessimism and we think oh my God nothing's going to work the last seven models didn't make it into production I feel like I should quit my job this AI thing is stupid I should just go back to doing something else maybe I could teach economics because I happen to know a lot about economics but as you're sort of like going through all of those things what are some of those things that with the with the power of the experience you've had you tell your younger self when it comes to working with all of these functions that quote unquote are there to help you deploy Ai and gen AI in a in a responsible fashion yeah um I think for you know a couple different themes there um you know there's been a couple projects that we've worked on that just ultimately failed in development right the business requirements and the security requir ments were just fundamentally incompatible there was no way that that um model was going to make it into production or that um that component was going to make it into production and uh it was a a really good test in Dev um and then just never made it to production and I think my younger self would have taken those a lot more seriously I would have been a lot more demoralized I would have been a lot more frustrated and at the time I think I was um but I think now having a couple successes of seeing either AI features or um you know AI oriented uh components move all the way to production it is possible it just takes a lot of time and negotiation and so I think having that longer view um has been helpful to me to know that you know when we are facing resistance or we are having those um debates you know that's part of the early creative process right to get this model to be the best it can be um uh those conversations have to happen and it might cause you to change some of the features um so you know for example we're um working on one application right now um where're we're using a large language model um that's you know fed ramp approved fed ramp authorized but there's some components of that large language model that we've turned off because we haven't had you know um full time to be able to assess and approve them or we just haven't gone through all those layers of approval yet um so we're still able to work with that component in our development environment and we're building a path to what we think is going to be productionizing it um but there's some compromises we made of um we need more time on this specific facet of this model um and so from yeah taking that Long View um it does take more time in government but um once you have a couple of wins you realize that it is possible you just have to you know work with your security Partners um and invite them to the table more frequently be be persistent don't give up absolutely so so last week uh on Thursday I happened to be in Meno Park New Jersey anyone know what Meno Park New Jersey is famous for light bulb light bulb Thomas Edison's team invented the light bulb there and how many tries did it take them a lot right I think some say a hundred some say a thousand you know somewhere maybe in in between there and and if you think about the role of people that build models what do we call them what is that position called what are they called they're called scientists they're called Data scientists right of all of the engineering of all the personas we have in the world of technology the one Persona that actually gets the ier of scientist is the person that gets to build models and maybe therein lies a clue that we're probably not going to get it right the first time and we're going to have to iterate and do it over and over again just as Jen said and be persistent sort of like that scientist that you're inquisitive you're not willing to give up you're stubborn and you're going to keep at it until you find a way to uh to to get it done and and meet the requirements and get to get to the right uh right outcomes um we I think we do have a few minutes for questions we don't have a few minutes one question one question okay uh one question make it really good you can pick from one of two topics North Korea or Russia just Jen okay we've got a question over here did anyone notice that we have Jen here talking about Jen AI yeah choke a lot right now do you get that get that every week yes I had a quick question on responsible AI approach and policy and how that's being worked or implemented yeah absolutely um so at the federal government level um extremely important extremely critical to the way that we work um so in October of 2023 you know the bid Administration released a really Landmark executive order on responsible AI um which really gave all of the federal agencies a a specific mandate around incorporating um responsible aif principles into our work you know at the development level um and uh really gave specific guidance around um models that specifically you know deal with human rights or safety impacting use cases um and then kind of reporting in governance that goes uh along with that um and we take that extremely seriously um so the responsible AI official sits within the center for analytics um you know they have stood up an entire governance mechanism um to review uh things that are being developed um and make sure that those principles are being you know upheld across our work at the state department that they're being um really infused as part of the governance process um it's not just one more thing to do but it's really part of the process of how you get something fully to production um so taking it very seriously um and uh you know really excited by the ways that um we're able to you know bring our security and responsible AI practitioners into our work um you know the center for Analytics um I suppose the last thing I would close with is um we started as the chief data uh organization and then uh like uh many of our peers became the chief data and I organization and so really all of that AI work Falls within um our organization um and so it's uh it it's critically important and one that we take take very seriously both in our work and for the whole department perfect well thank you Jen this was this was awesome a big round of applause for Jen thank you this is awesome the next panel session is going to be on data modernization in the era of AI uh so it's got Ethan Shen from FDA Cedar we have Brian Davis from data bricks moderating and Joe Dior from US Department of Veterans Affairs it's always tough to be the the uh session right after the break you just hope people come back so thank you um I do want to give a shout out to one

of my colleagues Lisa is running around thank you uh without without Lisa's guidance and pulling together content and speakers there would be no industry forum and one of her tasks is to keep me in line and uh so she sent me a slack yesterday and said make sure you print out the question was I read to mean don't go off script so so Lisa I have my script uh unfortunately Michael um called in sick today so um uh he will not be participating so um I I wanted to talk um we're to talk today about data modernization in the area era of geni I would say that um probably if you thought of government and modernization and Innovation those might be not the first three words that you know come to mind but I hope to um you know help to spell that rumor over the next 30 minutes or so um so I want to get started let Joe and Ethan introduce themselves um so please introduce yourself talk about your role within the agency and the work you've done with data brecks to this point I'll start with you Joe all right thanks Brian uh good afternoon everybody uh name is Joe Daria from US Department of Veteran Affairs I'm the director of data and analytics in the product engineering Service uh which is really kind of the core nexus of product delivery on U modern platforms uh in our office of information and Technology um our Marquee platform if you will our modernization Target is the summit data platform uh which is a Nexus of Health Data benefits data other administrative data sets uh where we are instantiating modern governance AI Readiness um all the really exciting things that uh you've heard about so far today uh I would like to say that you know VA is a large organization over 450,000 employees uh we're in the health care space we're in the benefits delivery space uh we're in run cemeteries right there's a lot going on uh at this department and one thing I will point out that uh one of my business partners here in the the audience said to me during the break um there our our DOD friends and partners at advana had a very interesting slide about their size and scale and there was one thing that they did not lead in and that was the number of doctors and health care facilities and and patients served and uh that would be our folks in VHA so just shout out to them only because you know some of them are in the audience and I want to get out of here alive um and then for everybody else uh that we've heard from this this afternoon uh in in the public sector area he thank you for your service and thank you for all the veterans we have in the room for your service lastly data bricks we use it it's [Laughter] [Laughter] okay yeah two stars yo um the uh in all seriousness uh data bricks really has been a great partner for us when we we started using uh Azure data breaks um I want to say it's like four four plus years ago uh in my area uh in the customer experience data warehouse they really helped us Bridge some feature gaps that we had with some other capabilities that were being rolled out into uh our government Cloud at the time uh and we learned about data breaks very quickly from that point forward and all the wonderful things we could do with it and when you know we we get AI executive orders and we get a focus on AI and everybody's using chat GPT and I think about all the things that we were doing in the ml space for the last several years prior to that land rush and gold gold rush Gold Rush whatever you want to call it land grab um data Brick's been a huge part of that for us and it's really been a catalyst for our team upskilling um and preparing for you know what was going to be next in the realm of data management analytics development and ultimately the realm of generative AI so happy to be here not two stars uh some of you may recognize Ethan he actually had a session on his program immediately prior to this which is a reminder all these Rec sessions are recorded So if you missed his session and you want to learn more specifically about what FDA is doing U I encourage you to go back and look at that but uh Ethan okay uh thank you and uh for those of you who attended my early session I'll repeat a little bit of what I said earlier about FDA so I'm Ean Chan I'm a division director of data management servic and solution under the office of business informatics in Cedar Center for drug evaluation research so which is one of the uh six seven six centers in the under FDA Food Drug Administration so we're focus on uh human drugs and uh make sure the uh safe and effective drugs is available to American public so that's our mission and we're about 6,000 uh strong out of 15,000 people agency nowhere compared to VA but we believe we have important Mission uh that to support uh equally important to support American public so I join FDA 2012 and uh there's a couple of things that happened in 2012 and uh number one things is there is a kadua act uh generic drug user fee Act passed in 2012 and which where we having a challenge as a nation that we want to make the drug affordable for American public where we having a lot of challenges annually we approve approximately 500 generic drugs at a time why we also have over a thousand applications in the backlog so that's where we were before we have the go for one the second thing happened is the Congress allowing FDA under the fidia 745 AA to mandate electronic submission so what does that mean that means before we having this one allowing FDA to collect data electronically FDA collect submissions in paper before 2012 that means what we today receive 400,000 submissions imagine amount of information we collect from paper each of the submission especially the original one is going to be shipped to FDA in trucks each of the sub has multiple copies we use yellow jacket blue jacket green jacket so we have a copy for records Management in paper we have a gigantic document room to hosting those paper jackets there so that's where we are 2012 2012 so after 5 years we enforce the Mandate of all the submissions for NDA andda and the bla biological license agree I mean application in electronic format and that's a mandate and in mandate and in 2019 that we I mean sorry 2018 we require commercial Rd to be submitted in electronic copy today we have about 99.97% as I check subm are in left TR format which means that will allow us to apply any of the mod Technologies we see in other words without those we're not datadriven decision making agency we're a paper driven decision making agency so huge difference that we're dealing with so as a result what we have today and I just checked the number we have approximately 900 50 generic applications being approved as of 2023 that's approv alterative approve which means we are making more affordable drugs available for American public so that's important for us now you go take a look how many drugs how many coopay you pay in your Pharmacy you probably will notice that and probably not I don't know right so so that's what we're doing here so hopefully that give you a sense of what we dealing with and that's the background what data and why data is important to agents say like us thank you Ethan so the similar question that was on the state and local panel and um it is uh where are you on

your data modernization journey and what are the lessons learned that you can share with some of the folks that maybe aren't quite as far along as you and I'll start with you Ethan okay well now I just mentioned about the importance of the data right so now now I have to talk about the data breaks so why data breaks so you see given the ability that we have electronic submitted data in a specific format that we see fit then we start to looking at abilities for how we be able to leverage the data we collect electronically so five years ago we're looking our ability before five years what we doing there is we have all those Silo system in what we called uh premarket approve process review process to the postmarket surveillance process each of them has multiple silos system people literally have to enter the data that they capture from the paper and Inter those system they all Silo they did not talk to each other there so the electronic data allow us to connect the information otherwise they're not being able to connect to our data standard allow us to extract the information how they been submit to us what we call the backbone file how we connect to them together so we started looking at the the technical ability how we be able to connect data there how we're going to be able to aggregate data how we be able to drive the Insight otherwise reside in the individual c system so we look at a solution to supporting star with the premarket regulatory review process to supportting that particular operation so certainly there's a choice of build or buy I know everybody dealing with that kind of situation there and and should we build ourself and from ground up there's lots of Open Source capability or should we buy the are mature cost product will be able to offer you quite a bit if not 100% quite a bit of functional that we're looking for so we look into a system that be able to scale up that's mature enough that has adequate support they're also looking forward not only being able to injust the structur and unstructured data but has a mindset to continue to deliver Advanced Analytical capability as a focus of the cloud-based Enterprise architecture so that's what we're looking for and we compare couple of products offers a similar Solution that's what we decide to go with the data breaks because we think that lak housee concept is the one what we're looking for and but we also in the same time equally worry about we don't walking into a data swamp as that time there people will worry about what is the future of the data Lake right so that's certainly something we worry about and uh but I think in five years we're very happy about the choice we made at that time there so today I did mention in my early session we do have twoth third of the cedar users are utilizing the Enterprise data platform that we established the decision we made five years ago we be able to having 14 different tenant spaces supporting different offices need because they're very diverse type of review process that the conducting cross life cycle of our review operation and we're also having 50 data sources being ingested internal FDA as well as external FDA it's nowhere compared to do 500 data sources that's a lot to us there we're not in their scale uh but I mean that's that's what we have we have about 70 use cases that we support as of today so we are able to having one what we call the One-Stop shop uh uh systems that allowing our review to be able to connect various information no matter where the original data is reside to then information is is readily available in front of the sky and I think it is important in a sense that what we seeing what we are able to achieve because in some of our review Community the kpi in their eyes to it system is measured about by the amount of Click they do to access the information they need so that's what we're dealing with and uh so I think there's quite a bit a journey that that we went through by adopting the system by adding additional capability I would say we're not quite there we only connected the information together there's a lot more s that we need to do so so I think where where we are in our journey is we've realized that we've gone through a lot of side quests to get where we were going and by that I mean when we we look at the scale and coverage of va just in the healthc care side of it with corporate data warehouse CDW roughly 7,000 workg group databases we've got 16 pedabytes of online storage from from the sand supporting at 700 plus SQL servers 10,000 SQL writers why am I saying all this I'm saying all this because it's big it's immense it's scale so how do you how do you collapse that into something that's modern Cloud native uh doesn't move data around for performance reasons right um so you've heard a lot about Medallion architecture you've heard a lot about the data Lake housee from from other folks today but we standardized on that concept architecturally uh because we wanted to bring compute to the data we didn't want to reproduce data assets in multiple locations for performance reasons we wanted it modern tooling we wanted people to be able to use things like data bricks if they wanted to we wanted people to be able to throw SQL at synapse and that's why we have that as an addition on top of uh top of data bricks so people can get to a more traditional look and feel and then we realized we had all these wonderful tools we to stitch them together in the data product concept um so in Summit we built uh the summit UI which allows us to stitch together all the third party tools that we have available to us uh to build this so where are we now right we're we're realizing that we're going to modernize CDW to Summit we've got pockets of other um work streams and and data silos that will be migrating to that platform but we are also building a you a fair amount of capability in what I'd like to call Federated governance and administration right what does that mean well we've got a data catalog of course you've heard a lot about that today but understanding where data lives what it represents what business process it connects to is the most important part of what you're going to do in modernizing your data processes and then building from from there uh how do you connect that right with with other uh with other silos how do you enforce Federated governance while you you start picking tools that are best to breed that tackle things like attribute based access control for you um things like we're using starbur for to even connect our Legacy on Prem systems up to Summit right so that we can Federate queries across those platforms so we're doing all of these things not because we necessarily want to reinforce the usage of these older platforms but we want to ease the transition right while we teach people how to use modern tooling keep them where they are keep them productive right if we don't have analysts if we don't have field practitioners I mean we have nurses that run SQL queries against CDW today to support Hospital operations right if we don't have a close analog for that user experience right and give them a journey not to overuse that word to the new platforms and and to the new ways of working then we're we're hurting veterans right we're hurting patient care we're hurting claims determination and administration we can't do that right so we have to balance these step functions of progress with ensuring that you still have something that people can look at and understand and say hey I know what that is I know how to deal with that and so you those 10,000 people that come off of CDW from a you know a SQL writer experience aren saying I don't know what to do with this Jupiter notebook right because that's what they say literally um and uh you know did you see Dr Scott's nodding knowingly he's like I've been working on this for five years I'm still not entirely clear is that with a why um anyway so the to give him a hard time he always I always call him out in every one of these things makes him feel loved anyway so so where where are we we're we're ensuring that we get the governance right uh we don't want to make bad decisions quicker we don't want to make decisions that we don't understand uh so data management whereas where we started data management is where you will be on every day in your AI Journey especially as you get to generative AI um and that's the message to take home get the get the stuff you've been trying to do right for five or 10 years get it right because if you're if you're going to move into the next phase without nailing that down you're you're creating a mountain of technical depth that you will never possibly be able to climb and you're you're going to set up the person who's going to do your job in a in a decade you know they'll be cursing your name like there's some people's names I won't utter to this day who maybe might have worked at VA I'm just kidding anyway hope that's hope that sums it up for you yeah well so kind of speaking of the

next step in the evolution right so you've built uh you've got a modern data strategy built on the lake house formation you've layered on governance and security and now here comes geni so um you know just generally speaking like what do you see as some of the most promising use cases for Gen in the VA I'll start with you Joe so if if you look at what our our partners in the CTO shop are working on today um in building out a governance framework and an evaluation guidelines for responsible AI many of the things that they're looking at aren't necessarily hey can we throw an llm at something or can we generate something quicker can we collectively get to knowledge right what what they're really looking at is how do we make end users lives better with AI um how do we get ambient scribes in the exam room with clinicians so that tiu or we call them TI encounter notes clinical notes you've heard that talked about a couple times I think if I remember correctly today um how do we get those generated quickly more accurately and how can we summarize them how can we extract clinical Concepts from them without a doctor having to uh to look at that there's Pilots around that going on uh I think focusing on not just what can we do with the data estate that we have or with generative AI but what can we do with the business processes that are out where we're serving veterans how can we assist that and I think the other thing that we're really going to find when we get serious about employing some of these Concepts I mean we've got a a rag pilot running on Summit today my AI lead Joe ratano is driving and it's really it's very similar to what you heard um I believe it was from LA county in terms of we've taken all of va's policy documents and Publications and uh putting something out there that we can start asking questions conversationally find out am I allowed to do this um you know how do I do this uh just as an example with our for our MLA Center of Excellence to say hey we did this once this is how we did it these are the tools we use these are the governance steps that we took to get there to ensure that the answers were accurate or at least representative and using that as a teaching mechanism right these very benign use cases that make people's lives easier um and becomes a teachable moment for people who are only used to you know getting chat GPT to produce uh song lyrics in the style of Britney Spears uh so that's where you gota you got to take people out of that mindset and say we're we're going to take va's Collective knowledge with what you know and we're going to do it responsibly right so we're not retraining models every six weeks and burning CPU Cycles GPU Cycles on that um because at the end of the day we're accountable for the dollars we spent um so as we start to bring these to bear and that's probably a common theme you've heard today in public sector if we bring these to Bear we will do so in such a way that is responsible ethical and governable thank you uh well on our side here and uh as you know we have a speaker tomorrow R is going to talk about what the hardware will be able to enable the capability that we envision I would say ever since open II announced the open its uh door for everybody to try there's a high hope so what FDA is not lack of as a scientist and uh what scientists like to do is exploring possib ities so we definitely have a lots of needs and the needs are mushrooming and uh uh that we're kind of trying our best trying to be able to find a way that'll be able to support them especially try to creating a Enterprise capability that will be able to support them instead of individually that what we need to do so that that's a challenge that we have in Enterprise level what we need to be able to do that also we also in the same time recognize there's no data model is perfect for every operations that you do and uh we will need to have different models there and we will need to have a platform that will allow us that first have the data as a base available for us to run the model and feeding the information to do the refinement to be able to derive the insight to support their operation two you will need to have enough resources that be able to sustain the computational needs of processing such data and so we know the hope is very high and we know the the resource demand is substantial and to a point there if we want to achieve and fulfilling all those need it's almost impossible for us to do it now further when we attend mosic AI yesterday it's even more scary and saying for the amount of resource you dedicated into that it's not guaranteed successful right I mean we all know we kind of come down as a partitioner we coming down to this floor there we are the one going to make it reality and that reality is not always Rosy and we're going to be the one to make it happen there what are the ones that will be able to make it more convincing and promising to support our operation it is a big question so I I'm saying that there's a lot of opportunities there's a lot of challenges and but we think having a platform to be able to support build up Enterprise capabilities important one we have a framework but we still have a lot of work to do uh so we're going to get the buzzer here in just a second but I just want to give get you out of here on one last question so short answer uh we are at the data and AI Summit obviously this is the high this industry form is everyone's highlight of the week but we still have two more days so what are you most interested what session what capability you most interested in learning about over the next two days I'll start with you E I think I'm here to learn from others from everybody's experience and practice to see how we be able to bring back to our environment to support what we need to do for our mission yeah very similar right there's a lot of great sessions to tomorrow and and Thursday that I'm signed up for that I can't wait to hear what other people are doing you know maybe one day I'll get you to to to do some of those things in some of our environments as we uh you know work with our partners to speed access to the uh the latest things that everybody's going to see this week um you know in our in our Azure environment all right really appreciate it thank you Ethan thank you Joe thank you thank you thanks thanks pleasure good

job all right um so for our our next speaker or our actually our next uh session will be with USCIS so it will actually be uh Customs IM immigration Service uh they were actually one of the First Federal customers for data bricks um actually have become one of our largest customers and actually did one of the largest PVC to E2 migrations have one of the largest un catalog um instantiations currently in the company so I think this should be a great session so Sean Peru uh welcome to the stage yeah excellent 25 minutes s all right hey everybody my name is Sean Benjamin um currently acting as a deputy chief of systems delivery but in my longtime role I've been with the data and business intelligence services at uais I've been with them since 2007 uh this project I've seen it go from an egg to a chicken to a flock of birds um it's it's been my uh life's project and uh it's not over it's still a growing um and I have Prov stress with me today he is our lead architect and engineer of pretty much everything data bricks um he is the man that works all hours of the day bringing this capability to us and all of our users all right I want to we're from USCIS so Customs immigration um a lot of times people we just say we're from immigration right so what that really means for us is you know we're visas naturalizations work permits uh visitors that's what we're working on that's our mission um and here I'm showing these are our director's priorities which line up great with the capability we can give her we really treat data as a product in our agency um we're about 15,000 users of the data and whole of our ecosystem um and we do make data driven decisions uh listening to everybody's speak today was so amazing because we all do share literally the same challenges little different flavors here and there but it's all the same use cases um and I'm going to really Center my presentation and put's presentation around not where we've come from to this day but really from about two years ago when uh Unity cat was first mentioned to me and I said I want that we've been chasing that forever so our first half of our presentation is really going to talk about what are the business cases what are we bringing to our business users what are we bringing to ourselves yeah we got to be selfish in this like I said Unity catalog once I heard that term uh and saw what that product could do I wanted it and uh through I talk to him every day saying hey we need this and we need this now I it's like how are we going to get this got to go to Ean and then you know my kind of what my long-term vision is it's not my long-term Vision it's already shared long-term Vision um with using data as a product and I'm going to hand it over to puru um and he's actually the guy that you really want to hear from um he he was in the trenches uh a lot of late hours a lot of long weekends and holidays uh bringing this capability to us so yeah right away starting off um what we I had to bring the users right I got to convince the users hey we're going to do another migration um I've already got all of you guys in data RS PVC we brought you a lot of capability but one of the things that we really wanted to show them is like hey every time there's a product drop it we can't get it because we're in PVC um so the only way we're going to get these new product drops and be in that testing working with data breaks you know as soon as something goes to a um you know like in a beta we can start testing it um and start planting our own sandboxes and bringing our users right in we are very much partnership with our users um primarily the chief data office which does not sit in oit um at our agency by the way it's uh Independence third part is a great relationship that way I suggest everybody set up their uh Chief data office that way um Delta sharing we already done it we have some uh friends here from CBP um where we uh created what my version of data mesh is and that's connecting uh my agency's data Lake to other agency's data laks um and we've done that through Delta share that's live in production um we actually had a use case the other day we had to bring some more data tables in because we built that pipeline through it took how many minutes to bring that data in five minutes five minutes five minutes whole new data set going to CBP um you know the people were part of it though I probably talk about two weeks for them to agree on letting us push the button uh Delta live tables uh SQL Warehouse uh we currently have a fairly substantial uh implementation of obie had that in our space since 2007 um about 2,000 users left on it now at one point in time 8,000 um I see SQL warehous is our way to helping us Sunset that product which I did announce this year on April 1st to 5,000 users I'm getting rid of obie um and they thought it was a joke and it wasn't I made sure April 1 was the perfect day Friday afternoon go home for the weekend come back to Monday full of emails um and obviously Unity catalog we've been chasing that we're all chasing that we don't know anything about our data we have 169 data sources and for us we we count data source as like full atto system and we just want to know how people are using the data we want to know who's using the data we have uh our fraud and uh National detection systems they want to know who's using their data they're building reports like how is our data being used by other people that's why Unity catalog has been so huge for me and data dictionaries is kind of it's not like a full data dictionary but it's going to give us a lot of capability I'm GNA probably be the only person to not talk about J today thank you um everybody else covered that for me and obviously governance and lineage I swear that that lineage is huge we build a lot of down steam prod s uh we have a couple large large data marks we're going to be able to watch that data travel all way through we're going to be able to find out how many versions of first names we have in all of our systems and so we can start working on the data governance um if om marel kicking her around his wife works for the data Chief data officer um and uh she works a lot with us on data governance and data management um and obviously cost efficiency everybody has always telling us like Sean how much is this going to cost us what does it look like and you know for me the price of this data this is the price of admission right to have your data fast timely quality because we really see our return on investment and actually our users time because that good data we we reduced our backlog at USCIS this past year for the first time in like five or six years by 15% which was a big deal our director runs around and tells everybody we want to know why because our director uses a dashboard that points to the same data everybody looks at you can't argue with that the director is looking at that data y'all better be looking at it too and we do again very data driven and I'm going to I see this this is really for me right when I talk to the business like all right they don't know what a DBU is it's so hard to explain a DBU to a business user they're like What's my per seat license no it doesn't work that way the more you play the more you pay and you just have this conversation I'm going to be able to show them that and it really helps us to tune our jobs it helps them to become again Partners in that data to help build that data product it's not come to it and ask us to build something for you it's come to it and ask us to help you build something we're going to get you moving and you're going to do it yourself um and then you can come and talk with us we'll help you along the way all right so that was the business

case now selfishly here's my case right I want to get out of that that Administration mode I don't need puru working at 2 o'clock on Christmas Eve he will anyway but I want to at least in my have some peace of mind on my side we're going to take that maintenance um give it all the data bricks uh huge Advantage we're moving as much of our products as we can into a software of a service software as a service model I'm obviously working with our security um and everybody else to to get that approved um the reliable efficient operations proactive monitoring that's where our really our job we decided we we need to be more platform based in the early days we were kind of experts in the data um because we could we could see all the data we could crosswalk it we could translate it once we brought in a CDO shop they started working on more data governance for us um they became intermediate intermediate you know with the business and us and helping them to understand the data we really took a heavy platform approach and part of that big heavy platform approach is that proactive monitoring um and then obviously the resource optimizations uh and a big thing on the bottom there um Unity catalog has been my Silver Bullet at USCIS to supporting our zero trust initiatives um that I've been working hand inand to to work with our security teams bringing this product in they've been actually helping us to support data breaks financially through zero trust initiatives because like I said it works for our business users it works for our security users um and we're definitely seeing major advantages of that uh two weeks from now we're actually going to be giving a couple presentations on the benefits of unity catalog for zero trust and we can do that because PR just helped us get into Unity catalog he'll be telling about that later oh too many too many all right the long-term Vision obious that future proof in uh I got chance to see a session on uh Unity catalog um manage tables today and I'm like that's it yeah this this is what we want yeah now I can get this because as soon as it's available to us we can start implementing it and pro is already like all right let's start some small test cases and you know what by the way any new data set we bring in let's make it a managed table right away so that way we can start working on the Technic technical debt plan to get everything that we already have that's not a managed table an unmanaged table right that's the right way to say it over to being managed but anything new we will be W man right away and again that perfect teaming obviously with unity catalog um of course I I said I wasn't going to talk about AIML or gen but we obviously know the advantages there that we're going to get um being in E2 there's no more waiting game um and we're pretty fast at USA that's pretty Progressive and I thought we were very fast even in the PVC but boy we're going to be way faster now um because those product drops they're mine um and then I can do what I do sell the business on these great ideas Pro we can do a does help Implement these great ideas um and the alignment of organizational goals and and optimizing data share and you heard me talk about Delta share um one of the big things doing that Delta share for that that what I'm calling the data mesh on our it's more of infrastructure data mesh working with other agencies um Jennifer from Department of State where are you you know like all these folks that we want to share data with there she is you know that's that's part of our tool and that's our thing but you know we need to start using it more internally we need to start using Delta share to broker that data between our Tableau instances which carries to most of our users Tableau and data bricks for us is pretty much peanut butter and jelly um you're probably having accounts in both places building a product visualizing it um and as well as uh connected to our SAS instances we have SAS um and connecting to all of our other Oracle interfaces we're really going to start taking advantage of the Delta share because you know we're going to have time now we're not going to be managing uh our systems late at night we're going to be actually Building Products in engineering all right so I'm going to hand it over to puru this is actually the guy you want to see not me I I talk to the executives he's the real Batman I don't know what I am probably more like the Joker so go ahead puru I'll get you started thank you son uh hi I'm purua I've been supporting USCIS for last nine years uh I'm from the uh contractor Contracting side of the house and currently IBM Employee uh to align with the usci vision of adopting emerging Technologies um improve uh architectural efficiency uh data governance and Zer zero trust implementation uh we took uh major like a initiative of migrating our datab environment to U der environment um this environment uh we like a as son mentioned he wants a unit catalog to get there we have to first migrate our PVC to E2 environment and subsequently uh to the E2 environment I mean the unity catalog so with that mind we also minimiz wanted to minimize as much as uh to user Community as seamless as possible so this is a strategy we have used and this is the different strategy stage is as you can see um uh most important stages is a u stage so we were able to if you look at this diagram it's like similar to a blue green approach so we could do a dry run up until U uh U stays to find if there is any errors if there is any issue and find out if there is any user impact so we are able to identify those with this approach and we could do multiple dry run to as seamless as possible uh this is the most important stage of our migration process for E2 environment uh this is a high level as Sean mentioned like he wants a Unity Cal to get there we started in 2022 uh at that time it was not even Federal um fed ramp uh we begin analyzing our in mark enement with the help of derix um so when we started we had at that point uh 67,000 hype tables we have uh uh roughly like 1,00 jobs workflow we have uh 65 databases connected by SAS we have a 12 subject area appointed by uh smart Obie we call it smart Obie we have over 3,000 data sour connect by a tableau so that was a lot of like our environment was first uh we have been in derik PBC since 201 uh 17 s 1 so we have already a lot of like a objects already in the PBC environment and this uh so we out uh we always have like uh uh in mind uh we going to be uh seamless as possible for user community so that's why um we in that previous slide stage was very important stage for us to understand what are the issue might face and what are the impact user may have um in this I will highlight couple of the there are so many challenges but I wanted to highlight couple of those uh one is like a at the same time um we are in the AWS our cloud provider AWS so we wanted to have a control plan connect to an derix AWS account we want to have this communication to go through a private link uh which we have created a VPC inpoint which basically our user web application to connect this use this vbc endpoint plus r API and secure cluster connection so it we don't want any of our traffic to go through a public internet so that was the one challenge we face in our environment um but we are able to uh connect those with the help of deric plus our usci networking as well as um different stakeholder DNS team proxies and things like that second uh challenge is um when in the PBC environment we we use uh derik's DHS uh I mean the DHS uh CFO certificate and load balancing in our account now when we migrate uh we have to use uh cloud. der.com which is a different URL that uh alone uh impacted all the user Community as you can think like all the user Community now they are they are need to connect uh new UR uh all the downstream app all the user base everyone is impacted so we have uh to come up with a lot of like a user um training a lot of communications everything we lifting and shifting to entirely new system all said and done uh we are able to migrate um these are the timeline uh it tooks a while but uh we are able to achieve uh we have a five different environment in our uh account and we are able to achieve uh last year before the D Victory is um so I would like to highlight few of the um uh once we migrated to E2 environment there are few I like to highlight uh one is as son mentioned like a um maintenance we used to we have a five environment whenever we have a PBC upgrade to new version it takes a lot of time and there are sometime we are ended up um erors and things like that and we have lot of support tickets and things like that so our time was spent on lot of maintaining those um environment PBC environment as soon as we migrated to the E2 environment there's no maintenance we don't have to spend those time we don't have to deal with the opening support ticket so all those time we're able to dedicate to something else some other prodive so we don't have to worry for lot of our uh support ticket went down drastically once we M to E2 environment another thing is we have seen um machine learning experiment um model training and uh registering those mod uh by our user community so they are really taking uh advantage of those autom ML and all those things um derik SQL is another one like uh we are able to utilize dat SQL as son mentioned we were uh we wanted to like our user base to Sunset some of the old Legacy application utilize start DB SQL we're also able to utilize uh this DB SQL for a lot of like a jdbc obbc connection as well as uh some other use case uh API things like that uh so that's was a highlight so we we didn't have that flexibility in PBC now we have all those new featur features I just wanted to highlight few of those I don't wanted to you know you all already know about the all the features uh um challenge faces I already went over um there are inrig all this was we are able to do in a u stage on the previous slide um downtown management so we learn everything while doing the dry run up until the you uh you know St stage uat stage so we we were able to minimize a downtime so we just took one weekend to migrate our production workflow so we took the long R that's why son always said like oh um so we are able to do a robust testing all those things in the UN States so once we migrate everything um then right after that we're start thinking of migrating to un catalog we we took same approach like this one we did exact same thing Unity catalog same approach we have different stages and we are doing similar to blue green approach so uh we started right after our E2 migration like a two or three months later we started our migration and um there are few challenges I like for example like uh when we come from the Legacy environment uh PBC environment to E2 uh one of the biggest challenge was uh the share access mode versus single access mode cluster um lot of functions and things like that are not supported in a share access mode whereas we are in the single access mode um but because of we have those uh stages and we're able to identify all the problem we may face in a uat stage and we do have a workaround so any Mission critical um we know how we can solve those some of the issue is like our language is not supported in the share access mode um machine learning is another one uh there are some uh p4j security issues and things like that so those are supported in single access mode but not supported on the share access mode that was one of the biggest uh issues we faced um and the I think uh once we identify once we have uh um workaround solution we're able to um deliver those another issue is like a volume um we in the E2 before we going to migrating to the unity catalog we are not able to create volume and deploy so we have to first migrate to the unity Cal then utilize the volume so that was another challenge but uh we knew like that's the way we have to go so we are able to overcome all those Challenge and successfully migrate to Unity catalog um so two this is the third week we migrated to successfully migrated to the unity Cal uh I just wanted to highlight few like uh the benefits of after migrating Unity Cog as son mentioned like a um you know we have a requirement to share the data to the external agency CVP and as soon as the requirement came we are able to deliver because of the we have the Delta share in unity Cal within the five minutes we are able to share those data to the external agency so this really makes uh we prior to that one we had open source like it take we did have a Delta share you know open open source Delta share but uh that initiative will take at least uh one Sprint which is a couple of week two we Trad two weeks for five minutes five minutes so that's one like ASA uh a second another like a as son mention lineage uh we're able to now see a lineage on okay this is upstream and this is a downstream and all those lineage were able to see um those that's that's capability we didn't have prior to Unity Cal implementation uh third is uh uh data governance every single objects in the unity catalog is uh access control and can be a anql statement Grant to blah blah blah to the your principal uh whether it's a user or service principal or um uh or groups so it every single is um you know access control and easy to implement those so that's uh one of the challenges we didn't have we we had before migrating to a Unity gal great job all right we're getting hook thank y'all thank much abolutely thank you uh some amazing customer stories uh I think as excited as we are about uh the announcements and the exciting progress in the field of AI um I think the most value users get is talking to other users and CIS has certainly been generous in sharing their Lessons Learned we know we've got a few other customers on the Custer migration with fed ramp High and il6 use them they're incredibly generous they've had great experience and they've just been a marquee adopter of unity I'm excited to announce the public sector

Awards uh I get to come up for the fun stuff um my name is Jude Bole along with suj mahanti we get to lead the uh public sector business um so oh questions and answers uh so I think first up is our uh the winner of our pubc partner of the year uh deoe GPS has been involved on numerous strategic projects across state and local uh fed DOD they've been taking complex technical challenges in delivering Mission and citizen impact and we're grateful for your help uh for the deoe team in the room we just stand up and be recognized and and partnership is so critical I mean in our sector maybe more than others we're all we're all working together uh and deo's been fantastic that way uh really making some Investments um we have a few finalists today for the data team Awards and those are going to be announced at the session uh this evening they'll be celebrated this evening so please uh if you're there make a point feel like we've got some good news maybe coming I I feel a little bit lucky about this board right here let me tell you a little bit about what they've accomplished this year so uh advana is the dod enterprise-wide uh Center for data AI it's multi-domain analytic and artificial intelligence platform as part of the chief digital and AI office cdao advana hosts dozens of business lines such as Logistics financial management Supply Chain management people and health advana puts the power of data analytics and AI in the pocket of every DOD analyst and decision maker the platform makes data decisions accessible understandable and useful to leaders analysts and Mission owners we also have another great story the LA County auditor and controller they collapsed 40 different systems uh across departments adopted secure scalable Lakehouse architecture increasing data insight and democratizing access they centralized data repositories built robust Vis visualizations for executive decision-making and streamline decision-making and streamline development of analytic tools this enabled realtime annual statistic reporting vendor payment reporting as well as comprehensive budget and expense reporting that previously required extensive manual efforts and time uh so real excited obviously about both of those stories there were other great stories that we wanted to share and celebrate uh some of our government customers have ethics limitations about where they can publicize and share and we all want to be responsible toward those but uh for those in the audience I thank you uh building something's not easy and a lot of you guys have the scar tissue to show um but come if you can to the data team Awards we're going to recognize uh some recipients tonight uh we do get to recognize though a public sector data transformation award and I'm excited to announce uh the recipient of that the 2024 industry transformation winner uh the advant team the advant team you you guys stand up come up actually damn it I want a picture with you where are you okay come on uh Cody Alex yes thank you so much Brad uh we've got a few other folks uh advana is a mission that touches a lot of different programs and it's just been a great adoption area for the Department of Defense developing AI uh and they were very generous sharing their insights today please one more time for the advant [Applause] [Applause] um now I realize I'm separating you from happy hour uh and you're probably ready for a drink um we have them downstair oh first we have some amazing sponsors um AWS who sponsoring the event you guys are welcome to step or you're welcome to have a drink with me we we'll see yall down there we'll meet you at the bar um thank you guys uh AWS and koft were uh two of our title sponsors we also have uh deoe who was one of our uh Platinum sponsors boo Allen and Accentra Federal Services were our premium sponsors I probably got the terms wrong but the information radiator doesn't have it so you don't know um but uh you know Aaron had once told me early in the process our leader of fed like you want to go fast go alone but if you want to go far you do it with partners and so we're deeply grateful for the partners who have helped us they you heard a lot of their stories today on stage with cdao uh with CIS uh with VA um and that partnership is critical so we want to thank you we particularly want to thank uh AWS and koft for sponsoring the happy hour which is going to be in the industry Forum area we provide you a map we try to make this simple that's us uh head to the mo Moscone expo hall uh this is starting in 4:40 so I'm not quite separating you from a drink but I will be in about a minute um we'll see you downstairs one of the most important things about this event is a chance to speak with other users to share their best practices and to benefit from their learnings you got a chance to hear some great stories today there are other great stories I see in this audience uh please make time to get a drink to share best practices and we'll see you throughout this week special thanks for sujan Lisa who put together a lot of this thank you very much guys oh [Music]