Skip to main content

Transforming Bio-Pharma Manufacturing: Eli Lilly's Data-Driven Journey With Databricks

Summary

  • Global Manufacturing Data Fabric: Eli Lilly established a Databricks-powered Global Manufacturing Data Fabric (GMDF) integrating data from various manufacturing systems into a unified data model for data-driven processes.
  • Serverless Architecture: The platform leverages Databricks Auto Loader for real-time data streaming, PySpark for automation, and Unity Catalog for governance, enabling seamless data processing across 16 manufacturing sites and 10 new sites coming online.
  • Use Case Delivery: GMDF delivers actionable insights across multiple use cases including batch release by exception, predictive maintenance, anomaly detection, and process optimization.

Transforming Bio-Pharma Manufacturing: Eli Lilly's Data-Driven Journey With Databricks

Watch: Transforming Bio-Pharma Manufacturing: Eli Lilly's Data-Driven Journey With Databricks
Talk by Abhijay Datta, Head of Delivery - HLS, Tredence ; SAUNAK DEBROY, Lead Innovation Architect - Data & AI, Eli Lilly ; Wilfred Mascarenhas, Executive Director, Eli Lilly and Company
Eli Lilly and Company, a leading bio-pharma company, is revolutionizing manufacturing with next-gen fully digital sites. Lilly and Tredence have partnered to establish a Databricks-powered Global Manufacturing Data Fabric (GMDF), laying the groundwork for transformative data products used by various personas at sites and globally. By integrating data from various manufacturing systems into a unified data model, GMDF has delivered actionable insights across several use cases such as batch release by exception, predictive maintenance, anomaly detection, process optimization and more. Our serverless architecture leverages Databricks Auto Loader for real-time data streaming, PySpark for automation and Unity Catalog for governance, ensuring seamless data processing and optimization. This platform is the foundation for data driven processes, self-service analytics, AI and more. this video will provide details on the data architecture and strategy and share a few use cases delivered.
Here's more to explore:
Databricks named a leader in the 2024 Gartner® Magic Quadrant™for Cloud DBMS: https://www.databricks.com/resources/analyst-paper/databricks-named-leader-by-gartner
An open, unified approach to your data, BI and AI workloads: https://www.databricks.com/product/databricks-sql
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

Chapters

FAQs

What is Eli Lilly's Global Manufacturing Data Fabric?

GMDF is a Databricks-powered data platform that integrates data from various manufacturing systems into a unified data model. It serves as the foundation for data-driven processes, self-service analytics, and AI across Eli Lilly's manufacturing operations globally.

How does Eli Lilly use the GMDF for manufacturing insights?

GMDF delivers actionable insights across use cases including batch release by exception, predictive maintenance, anomaly detection, and process optimization. These insights help drive operational efficiency and decision-making across manufacturing sites.

What technical architecture powers the GMDF?

The serverless architecture leverages Databricks Auto Loader for real-time data streaming from manufacturing systems, PySpark for automation and data processing, and Unity Catalog for governance. This combination ensures seamless data processing and optimization.

How does Eli Lilly's growth strategy benefit from GMDF?

Eli Lilly has 16 manufacturing sites today and is building 10 new sites to come online in the next 2-5 years. The GMDF's scalable architecture positions the company to be ready for this anticipated growth in manufacturing capacity.

Full transcript

0:00 Introduction and company overview

All right, good afternoon everyone. Thank you for being here. Um, I know this is 1 hour after lunch just when the food starts kicking in. Uh, food coma starts kicking in. So, we'll do our best not to uh to keep you awake. Uh, so the topic of for today is uh datadriven journey with uh with data bricks. Uh. I'm a technology leader and the chief architect in manufacturing and quality uh primarily responsible for uh innovation and strategy.
And then we'll have and uh Abijj who are co-presenters uh introduce themselves um when they're when their portion comes up. So um for the agenda we'll start with um a little bit about background on our data journey. So I'll do that. Following that, will uh present on our architecture. We'll try to get a little bit deep into the architecture. That way you guys can take back something um from our architecture or from our implementation.
Um and then ABJ will share key learnings from they were our partners from Trance. Uh they'll share some key learnings uh bringing outside in perspective um and helping us in our journey to implement uh data bricks at. and then we'll have some a few ca use case highlights uh and a demo and then I'll also talk a little bit about what's next. Okay. Uh so a little bit about. Um we unite uh caring and dis with discovery to bring life-changing medicines to patients around the world.
Um and everything we do is uh is based on our core values of integrity uh excellence and respect for people. Uh we are based in Indianapolis, Indiana, headquartered there. Uh but we do have um sites both um sales and marketing offices as well as um um manufacturing sites across nine countries. So for this presentation what is more important is the um is relevant is more of the manufacturing sites we have across nine countries.
We are today around 16 manufacturing sites and another building another 10 in the next uh that should come online in the next 2 to 5 years. So it's a tremendous uh growth and anticipated growth and the kind of what we are going to share today u you know lends itself in terms of being ready for the anticipated growth in manufacturing. So like I said, you know, is going through an amazing uh period of growth. I think this is um been um I I think an unprecedented growth uh not only in in our industry on our history, but I would say also amongst all of our pharma peers.
We've got three molecules right now that are in uh late in regulatory review. We have 22 in uh phase three, 24 in phase 2 and 26 in phase one. So, and we are hiring very aggressively. So, I think just in it, I would say in the last 3 years, we've doubled our um our uh resource or headcount. So, if you guys know of anyone who's looking for a job, you know, do encourage them. We need the best talent we can to bring, you know, again, life-changing medicines to patients um across the globe.
Um lily. com/careers in case you guys um so so in terms of our data journey, you know, this really started with our leaders um being very intentional in terms of our IT stack. We've done a great job of ensuring you know all of our manufacturing sites have the exact same IT stack across across the globe. So our uh ERP systems, our quality management system, document management system, manufacturing execution system are pretty much the same across you know all of our manufacturing sites.

3:42 Data journey and organizational structure

Even if we have acquired those sites, you know, we've been very disciplined in terms of making sure um that they have the common IT stack and that really has given us a huge advantage in terms of harmonizing data and bringing all that data into a single data store uh which we again have qualified at the GMP level for those you know in non-infarma GMP stands for global good manufacturing practice which is a requirement for us in terms of uh using data for regulatory purposes.
So really you know our our journey started with harmonizing the application stack which again helps us harmonize on the data uh create a centralized data store and then bring that data uh for to drive value to our to our business. So again the D journey started in you know 20 years ago. Um as we started harmonizing on the application stack and started implementing the application stack we were also very disciplined in terms of building the global data warehouse.
So we started with the data warehouse using the traditional um means of you know the structured data and the data warehouse concepts. Following that, there was a again a big increase in the use of data building dashboards and our uh our business users wanting to do more and more self-service BI. So we essentially um improved that architecture, created a high performance uh built it on a high performance appliance. Uh following that you know there was a big push for unstructured data uh looking to analyze um unstructured data whether that's videos um audio or most more importantly unstructured text right and then we had the the whole phase of adoption adoption of Hadoop uh adding to uh our data strategy and then two years ago we um essentially went to the next uh improvement ment in our data architecture in modernizing our data architecture which is leveraging the lakehouse concept uh and leveraging again data bricks as our implementation platform uh for the uh lakehouse and again um in addition to the technology work we are doing in terms of building the lakehouse we're also being very intentional now in terms of building the data products uh the product data products along with you know data governance ensuring those data uh products are appropriately stewarded by our our business um so on and so forth and then we that has again helped us um add new concepts like AI uh and then chatbi on our data products so next what I'll all right so um in addition to the technology so we'll talk a lot about technology in this presentation but I also wanted to point out that there are many other key ingredients to for towards our success in in our data journey.
So um in addition to the innovation strategy work that my team does, we also have uh a team that uh does data governance uh and that's led by uh uh you know focusing on mastering data metadata uh harmonizing data data catalog data um you know the whole aspects of data governance and then we also have uh Jasm Singh who leads our uh scale up and value delivery uh team uh ensuring you know what we what we do in in innovation can be scaled up and performs to deliver again uh value to our uh end users.
And then the other other thing we've focused a lot on uh again thanks to uh data bricks and then uh partnership with treadins is ensuring that we can bring our um our business as well as IT folks um up to speed and in terms of upskilling. So we've you know one of the strategy we use for this is doing hackathons. Uh we did one a couple of months ago and we are getting ready to do another one uh in in July. Uh and what we do there is you know encourage the business and IT folks to bring in new use cases and then we've done one just using Genie saying okay you know this is your use case. uh we have a a genie room where we we uh enable the data products that they need to use and then we ask them to solve the problem just using Genie.
So that this has been an amazing uh success the last one we did and we're looking forward to again you know um upskilling even more people uh in the coming uh coming months. So with that I'll let uh I'll hand it over to who's going to talk a little bit about the architecture uh and following that tradance will talk about the um the journey. Thanks. Myself Shaak Debra I'm part of the innovation team in uh primarily responsible for delivering data AI and digital tech solutions at. uh in this slide I'm going to cover more about our our highle design flow what we have built and then in the next slide I'm going to touch base a bit in depth in architecture covering about our core architecture components and the services that we have used so on this on the left hand side the mark is not but on the pretty much on the left hand side we have this various source systems the lower two blocks are primarily our IT systems which are mostly our enterprise IT and our manufacturing IT and then we have the top two blocks on our OT systems.

9:01 Architecture and technical approach

The moving on through the right hand side we have the various injection frameworks we have built and I'm going to touch base upon that in in my next slide on the detail architecture. Uh but primarily the idea here is to get the data into our lakehouse platform in the integration of our IT and OT data into the data platform. So here on the on the middle layer is our lakehouse or the data platform where we have pretty much built our medallion architecture uh where we have the raw refine and the confirm layer pretty much staying true to our bronze, silver and gold layer.
So confirm layer is the final layer where we build all our data products and then we have unity catalog doing all the data governance and the data lineage and the unified uh data. So that's pretty much our data platform and then we have consumption uh we do traditional reporting analytics and then we have AI and ML use cases which is being served through this data platform and we do expose this data products to the marketplace.
So we have like an well- definfined access strategy and then we have uh data product owners who then have who own the end toend ownership of the delivery of the data products and the life cycle management of the data products. So in summary what I would like to say here is like what we heard today in the keynote we have tried to say stay true to the to that architecture. So we have the lakehouse uh platform in middle and then we have unity catalog on top of it doing more of the unified governance and then we have a IML and the reporting in terms of doing the consumption and making the platform ready for AI.
With this I will go into the detail architecture. I know this slide is bit busy. I will try my best to keep it simple but yeah. So on the very left hand side we have uh the IT data sources which says AWS EDB. Uh it's primarily our IT data sources and as part of our enterprise strategy we store the data in AWS. We do have built like an custom injection component to get the data into datab bricks. uh we are working very closely uh with uh datab bricks leveraging lakehouse federation to kind of do that external movement into the data or external cataloging into the data and so that we use some of out of the out of the uh out of the box component to get the data into cloud on top we have our manufacturing IT data and it's pretty much the traditional uh change data capture injection which we do through Azure data factory uh on top uh we have our OT data which is the PI historian. uh we have all our instruments and equipments data and this is one place where we invested a lot of time to build kind of a custom injection component to get the data from our legacy PI historian into into cloud but having said that we are again also working very close with our vendor partners like Aviva Pi and datab bricks to ensure that we get some out of the box integration components so that we uh we again stay true to getting the out of the box integration components uh then on the On the top we have more of our standalone.
We have few standalone instruments which are like pH meters, weighing balances and we use our IoT and MQTT infrastructure to get the data into cloud. So in summary uh we have two principles which stands out here. We have our IT data which is like the batch injection which we do and then we have our OT data which we do in near near realtime streaming uh into the data bricks cloud environment. And the second principle which we have tried to stay true is we work as much as possible with data bricks and our vendor partners to do out of the box integration.
So we don't do uh custom we do have to do we are still doing some of those but we still work very closely with them to ensure that we do out of the in out of the box integrations as we go on the right we do have like the raw layer refine layer and the confirm layer which I spoke in my earlier slide. So you have the data moves from raw to refine where you do the traditional change data capture merge and then the data moves into confirm layer where we build our final data products for our consumption.
On the top is the datab bricks unity catalog which pretty much does all the access governance all the security control the lineage and all of that. So as we move the data from raw to refined to confirm we have uh unity catalog unity catalog doing all of this uh unified governance. Once the data products are built we have our consumption patterns on the rightmost and then we have powerbi which is uh traditionally used for all our traditional reporting and analytics. uh we have one lake which we are leveraging again we have done successful integrations with Microsoft fabric and as fabric is based uh built on the same lakehouse concept we are very easily able to integrate that into one lake and that has been a gamecher for us where now site engineers can have their fabric workspaces and do many of their uh self-service analytics we have genie on the top and I'm going to do a demo at the end where we we going to see some of the self-service analytics like NL LQ on top of uh on top of your data products and as we do this we we do have like we are also positioning this platform to make this AI ready.
So I think is going to uh touch base upon that at the at the end and summarize on that. So yeah with that I would hand it over to uh who is going to talk about our partnership uh in building this entire ecosystem. Excellent. Is this working? Thanks. Thanks, Ronuk. Uh I hope I'm audible. Yes. Okay. All right. So, um so just a little bit about my introduction. So I head the uh solutions delivery for our healthcare business in Trance.
Um before we get into the journey, a little bit about Tredence. So this is a company which is focused on data and AI solutions development. Uh headquartered in San Jose. Uh we work for several industries. So we are deeply also invested in healthcare and life sciences as well as manufacturing and uh supply chain domains. Um so we are fortunate enough to be involved in transformation journeys for several companies. But the one thing and I'll of course as I as I kind of talk about the transformation journey the one thing that I would say is the biggest learning for all these transformations and especially the ones that have been happening on the data space is while there is technical uh complexities that you have to overcome but there is that's only one side of the coin right so we you have to have a mindset of driving business outcomes and you cannot leave it too late you cannot have all the nuts and bolts all built out and then start driving business outcomes because they get pretty impatient and that's the journey I think when we started working with that was uh that was within the with in the DNA of the transformation from day one and that's the that's the learning that I wanted to share with this team that how do you kind of chart this journey how did it and um from day one right this this whole journey started 18 months back or maybe a little bit more than that right and from day one it was never a vendor client relationship, right?

15:47 Partnership and collaborative transformation

So, has been uh very accommodating. They they wanted to also listen to ideas outside in they wanted they wanted to also see how we have done these uh similar transformations elsewhere. what kind of learnings we can bring in. So if you look at the if you look at the journey right so it started all with the vision setting with business as Wilfrit was saying there there are a lot of site users even though the tech stacks and the application stacks were common but it was it was still disparate right the sites were using different solutions in the sense like they were using it differently they had different needs the businesses are very different and as business was scaling they went about and started under I mean of course the need for modernization was there but they also collected the business requirements they understand understood the high level pain points and the road map that they put together was a data productdriven road map uh which which again is a pretty much a game-changing um uh uh a gamechanging uh uh aspect that they brought in.
The reason I say that is because if you look at how the solution evol evolved the data products is a concept where you bring in very highly curated data products on specific aspects of the domain. They are owned by data product owners. The attributes which are which are there in the data product they lend themselves to a multitude of use cases. Right? It's very tightly governed. I think when he was speaking u he also mentioned about the u his um counterparts in the governance team Diadra who also manages the overall governance of this data product what comes in how do you manage the quality who uses it the permissions and so on and so forth.
So the first few months of our journey went in in identifying and prioritizing the right data products. So we went and developed the data products the uh we we built as much as possible in the first few months but as I said right I mean you have to have an eye for business outcomes and you cannot just keep building on a data product journey forever right so we we created two more work streams on top of the data product workstream and one what shauk was covering was the overall platform hardening right which you see and I don't know whether it's it's in the middle in the bottom so the whole data platform that we were building Right?
While we wanted speed but we also didn't want to lose sight that the platform has to be built in the right way. We have to build the right CI/CD. We have to as as was also showing the governed by Unity catalog. All the pipelines that we were building it was biodal. So there were pipelines coming through real time. There were pipelines coming through batch. How do you harmonize them? as much as possible make it flexible, configurable so that you externalize a lot of the logic outside and you do not have to keep coding so that you do not get into a lot of technical debt in this new platform from day one.
So that was the first work stream you do data platform hardening and then the second work stream was how do you build solutions for end users because the data products were coming up right though it was not fully built out all data products were not built out so there were a couple of data products that we started with and with those two data products very high quality data products we started thinking what solutions we can bring to the end users again uh we when we started this journey we embedded ourselves in the manufacturing the DIA teams of and we got a lot of help to get connected with the business users understand what they're looking for and the mindset was about agile experimentation and building new solutions right so it's not about spending months in understanding what the requirements are it's about building something quick get in front of the users show them get them excited some solutions may work some may not right so the solutions which work you go for scaling which is the last last thing that you see which is which is where uh when showing a the orc chart just meets team came in solutions which passed through the filters which went for a full full-fledged scaling with the sites um and when I when I kind of there are very specific aspects of this transformation that I wanted to share and this is something that it's almost like a playbook and uh and I don't know whether it's readable from but I'll anyway talk about it so there are very few things which are very critical to the transformation to ensure that this does not become a technical rat hole where everybody just focuses on technology how to bring in data and uh you are you continue to ingest data in this big platform and there is very little sight sight of the end users right so when we started working with it was very clear to us that we have to start thinking about business outcome right and uh so alignment with the business goals from day one was the mantra that brought in and we also kind of when we shared some of the learnings with them.

21:24 Key learnings and best practices

So some because we also were embedded with them we had some domain aspects that we brought in. So we started working with the end users um and we got the requirements very very clearly laid out so that we knew what data products to be built and what solutions to be built. The data quality and lineage this is a very critical thing right. So a lot of data in the source systems they are not in good shape right and we know that right the data the source systems they they are weighed down by decades of legacy processes may have moved out of source systems.
So how do you bring the data in? How do you curate the data? How do you do the right kind of data quality treatment on that data? Again, this is this is something where domain technology all get mixed together. You use AI in in parts of the data quality treatment. The big thing that I would like to talk about in the data quality when it comes to the platform for those of us who are in this journey right is a large part of the data quality has to be configurable has to be externalized um so that you know you you are not constantly developing new quality rules.
You have to you have to understand the framework. You have to build that framework so that as more and more data products get built, more and more data sources get ingested. It it is it it doesn't take too long to build out the quality rules and the transformation rules. The third thing again I I again this is this a big reiteration of the mindset of uh and we we are part of many many transformations but this is very unique to is how do you how do you have this fail fast approach? uh of course the um fang companies and many others they they revel in this right they they bring products in the market it doesn't work they move on but it's very hard for an enterprise like to actually embibe this mindset right and we have seen this firsthand as the data products started getting built we we actually collaborated with the business and we often would not know what we are building is probably the right solution or not so we went ahead and built some of these um some Some actually went for scaling.
Some we actually paused right when we saw that it probably is not meeting the business needs. Data product thinking I've already talked about it. Um it's it's it's a I think for this has been the biggest benefit because the way the data products are developed and existing data products the it gets it keeps getting enhanced. Then there are new data products in the pipeline. They're clear owners of these data products.
These are very very high quality data getting sourced. I mean data is sourced from source system but the treatment of that data and putting it into data product with clear owners who are directly um connected with the end users to bring in the right attributes bring in the right modeling so that all kinds of solutions data science AI use cases can be built on top of it. And the last one is very important is the change management and user adoption. never easy when um you are you're operating at the scale of right so you can do all the pilots with one site with a small set of users but once you start scaling once you start taking it to other sites it just it does get difficult right so how do you find the right kind of business champions who take the solution who own it who kind of take responsibility to bring it into their sites into their sphere of influence and ensure that the solution gets gets adopted the business outcomes gets driven is again something that we have found extremely well done within uh within and a lot of these things have been a very shared partnership right so we have been as I said embedded teams that we formed throughout this transformation where integrated teams so uh uh uh team our team we all got embedded within a u um development team or business requirements gather ing team and we collectively work towards the outcome.
So we were also as much vested in it to ensure that the outcomes are the outcomes are there and business is really adopting these solutions. Again I can go on for long enough. I'm pretty sure I'm missing a lot of the learnings but these are some of the very key things that I wanted to highlight as part of the transformation journey. So I want to hand back to show you some solutions I think uh which we brought to life with this.
Okay. So um I'm going to cover here like a genie space and use case demo which we build. So as we all know Genie is like more of your natural language query on top of your data products or your data sets like so yeah so in this demo we just we are just going to showcase here like it's it's our datab bricks workspace we have like created this genie space uh and essentially just all once you bring all your data so we have our asset data product and we got all the desired data sets on the on the Genie space and then you start asking questions on top of it.

26:34 Genie workspace demo

So here we are asking like show me all the work orders for a a given asset and you can pause it. Yeah. So as you can see here we are uh showing like ask give me all the work orders work order number description status for the asset number. We have masked some of those uh for security reasons but yeah uh that's how it uh looks like. And then when we give the asset number and it it just does give me uh the information in the tabular format right uh it pretty much again uh gives me the SQL code the that got generated.
So today what we were uh seeing in the keynote it pretty much does it very nicely. So and it's all built on top of your Unity catalog. So one great takeaway from this is like as you build your unity catalog and how you define your metadata and how you define your data set is primarily responsible for how Genie performs on top of your data sets and gives back the query. We have been testing this and we have now make this prime ready for production.
So we have this workspace and the Genie workspace for on top of our data products which are essentially then released to our users who are then just asking questions on top of it. So here are some uh other examples which again uh features around asset data product but essentially this is just to showcase like it does not just does not stop at one uh you can keep on asking questions and it does give on the iterative loop.
So it does give you like an conversational chat. So you just give you the next questions that you should be asking. Uh so yeah here we are asking about what is the average duration of work orders by work type and then it just gives you the data and then it also spills out some uh beautiful visualizations there. So you can uh download this and take it away. So that's the self-service analytics vision which we had right.
So you essentially create your data products or data sets in the unity catalog and then just expose that to a genie space and you have your end users then asking questions who are the real end users who know the value of the data who who can then use the data and you don't have to have like an SQL background or any code background and behind the scenes it does it very uh I would say very seamlessly here. Um so yeah and as you can see this was a pretty uh complex code and here is the genie room which here you can get all the data sets.
It shows you all the past chats. So all the chats that have been happening people have been asking questions. So it does give you a good uh information. It does give you the history of about how the users are asking. So if you have asked questions in like past 10 days back you want to go and visit that you can go into the history. You can check that. It just give it also gives you a rating uh so you can like do like an positive and negative and that feedback in the loop does works.
So with that uh I would say I would give it to showcase more of our vision in terms of how do we now position this platform as part of our AI ready. Okay. Thank you SA. So um we intentionally selected the Genie demo uh because again the you know the future is I in my opinion at least more chat based um querying and then also more you know I I believe the dashboards of tomorrow will be more chat based rather than you know building pre pre-made dashboards and then giving those to the business.

30:21 Future vision and Q&A

So uh we will continue to do that. We're still doing that but again this is something new that we've we've been partnering very closely uh with data bricks to bring to the to the end user. So in my opinion you know finally we may have actual self-service BI uh be realized uh across all of uh all across the enterprise. So to summarize again we talked about you know building very good data products that's kind of the foundation of uh everything we are doing.
On top of that, you know, we're ensuring we have the right governance. Um, making sure that, you know, the master data is is common across the data products, the reference data is there, the data lineage u appropriate stewardship of the business, so on and so forth. And really that is what is going to empower us uh in terms of bringing AI and not really perhaps in you know in the future not even creating data products but where the data products are created on the fly and that I think in my opinion will probably be because of the discipline in in mastering data in in ref having the right reference data in creating the right ontologies and you know knowledge graphs and so on and so forth where the data products of tomorrow would be built by AI versus you know having to build these data products uh targeted to each and every business problem.
So that's kind of what we are looking at today. How do we make our data products AI ready? Um and there's a lot of um you know a lot of attention to this space in industry. We see saw that this morning in the in the keynote. Um and the specific things we've started working on right now is creating these ontologies and knowledge graphs. Right? How does data connect with each other? So that again as you are writing these queries uh using Genie or other things the integration of data how data relates to each other is kind of pre-built through these anthologies and knowledge graphs.
So then you know you there's little um variation or little possibilities of hallucinations or getting wrong results etc or having the need for a human in the loop. Uh so we're spending a lot of time right now uh we've started this journey of building the knowledge graphs and ontologies. The other thing we're um also looking at is uh creating agents on top of these uh data uh data products. So creating targeted agents for each of those data products and we we heard this morning uh you know data bricks doing the agent bricks.
Uh so we're really really excited to see you know how we could potentially leverage the concept of agent bricks and bring that on top of our data products. The other thing is also you know vector DB is the ability to query your data especially unstructured data. We have a lot of this interest in this space and we've um now have a lot of people you know building their own vector DBs. So we're also trying to harmonize on that creating a centralized vector database so that again it has the proper chunking strategies embeddings etc. so that again each business user doesn't have to do that for themselves.
Um and then additionally we're also looking at some synthetic data. Uh we're investing in some tools in in adding uh synthetic data to our data products. Again as you need more data for AI um you know we want to kind of harmonize that along with the data product so that again you don't need to create that synthetic data for each and every uh every project and there are lots of other concepts we're looking at data labeling um so on and so forth so that again these are aspects new aspects to consider as you build the data products um of tomorrow the AI ready data products of tomorrow.
So that's really all we had. Um hopefully that was that gives you some glimpse into the work we are doing and some best practices we are following and then you know what's coming up next. We have five minutes for Q&A. Um so happy to take some Q&A questions.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.