Modernize Analytics and Empower AI in Databricks with Trusted Data Using Informatica

Summary

• Enterprise Data Management Integration: Informatica's Intelligent Data Management Cloud (IDMC) empowers analytics and AI on the Databricks Data and AI platform by enabling no-code/low-code data pipelines at enterprise scale.
• Data Quality and Governance: Informatica extends enterprise governance with Cloud Data Governance and Catalog (CDGC) integration with Unity Catalog, ensuring clean, curated, and governed data for analytics and AI initiatives.
• Accelerated Implementation: The partnership with Databricks enables teams to accelerate pilot-to-production timelines while improving data quality through Informatica's data management capabilities and Mosaic AI integration.

Modernize Analytics and Empower AI in Databricks with Trusted Data Using Informatica

Talk by Ajay GOLLAPALLI, Director, Informatica (Sponsor Speaker); Rik Tamm-Daniels, GVP Ecosystems and Technology, Informatica (Sponsor Speaker)
As enterprises continue their journey to the cloud, data warehouse and data management modernization is essential to optimize analytics and drive business outcomes. Minimizing modernization timelines is important for reducing risk and shortening time to value – and ensuring enterprise data is clean, curated and governed is imperative to enable analytics and AI initiatives. In this video, learn how Informatica's Intelligent Data Management Cloud (IDMC) empowers analytics and AI on Databricks by helping data teams: · Develop no-code/low-code data pipelines that ingest, transform and clean data at enterprise scale · Improve data quality and extend enterprise governance with Informatica Cloud Data Governance and Catalog (CDGC) and Unity Catalog · Accelerate pilot-to-production with Mosaic AI
Here's more to explore:
The best data warehouse is a lakehouse: https://www.databricks.com/product/databricks-sql
How to migrate legacy data warehouses: https://www.databricks.com/resources/ebook/migrate-your-legacy-data-warehouse-databricks
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

FAQs

What does Informatica's IDMC provide for Databricks?

Informatica Intelligent Data Management Cloud (IDMC) enables no-code/low-code data pipelines for ingestion, transformation, and data cleaning at enterprise scale. It integrates with the Databricks Data and AI platform to support analytics and AI initiatives with trusted, quality data.

How does Informatica improve data governance on Databricks?

Informatica Cloud Data Governance and Catalog (CDGC) extends enterprise governance by integrating with Databricks' Unity Catalog. This combination ensures consistent governance, improves data quality, and provides visibility into data lineage across the organization.

What data management challenges does Informatica address?

Informatica helps organizations identify where data lives across hundreds or thousands of systems, extract data efficiently from various sources, and ensure data quality and governance. It provides wide connectivity to diverse data sources with best-in-class connectivity to Databricks.

How does the Informatica-Databricks partnership accelerate time-to-value?

By combining Informatica's data management and quality capabilities with Databricks' analytics and AI platform, organizations can accelerate pilot-to-production timelines. Integration with Mosaic AI further enhances the ability to build trusted data foundations quickly.

Full transcript

All right. Well, thanks so much everyone. I know the uh the after lunch slump can be a little difficult, but I appreciate you all taking the time to join us and hopefully we'll have some more folks uh trickling in as we go on. So, today we're going to be talking a lot about how Informatica empowers an enterprisegrade foundation for data bricks for both analytics as well as AI use cases. Since we are a publicly traded company, I do have to put up this forward-looking statement.
So, uh, safe harbor and that sort of thing. And then a quick reminder, uh, it's really helpful to us when you guys complete your surveys after the session and give us feedback on the content. We obviously want to make sure you guys get a lot out of this. Um, just to get to know the room a little bit. How many of you are current Informatica customers? By just a show of hands. One. Okay, good. So, a lot of new folks. Um, and are two folks.
Are you guys on Informatica's cloud technology or some of our older on premises technology? the old onrem down the on-rem too. Okay, great. All right. Well, so the great news is we're going to talk all about our cloud technology which is what we're 100% focused on as Informatica today. So we think about Informatica and data bricks. I always like to think in terms of what are the core data management challenges that Informatica is aiming to help address.
And we think about your typical data journey to data bricks. There's often a big challenge especially for a big enterprise you know identifying where does my data even live right I have you know hundreds or thousands of systems uh when I want to bring data to data bricks different use cases how do I even begin to make sense of my existing landscape to find the right information that's kind of a data cataloging type challenge the next step is once you're able to find your data assets then it's okay can I extract them from where they live do I have the right connectivity and can I get them to data bricks in the most efficient performant way performant way possible and this is a big strength of Informatica for a long time on connectivity to a wide range of data sources birectional connectivity but we also have a very deep partnership with data bricks a direct engineering toengineering collaboration that ensures that when you use connectors from Informatica you're getting best-in-class connectivity to data bricks the next challenge often becomes well great I found my data I got it out of where it is I got it to data bricks but is that data trustworthy is it usable or do I have data quality issues I need to identify and remediate or do I have a situation where I've got sensitive data and I want to make sure my data access policies and rules across my entire enterprise are consistent but they're enforced within within data bricks.
I want to look at in some cases uh that data quality and cleansing extends all the way what we call master data management which I'd like to think of master data as the is when your for cases when your data can't afford to be wrong. Supply chain is a great example, right? If your supplier data is incorrect, it could be millions and millions of dollars of impact to you. So master data really is about taking a lot of complex, disparit, overlapping, conflicting data and merging into a single version of the truth, right?
And last but not least is data governance. And data governance really does become the engine of data democratization is how we look at it. And that governance really keeps track of where data came from, its provenence, its lineage, but also what does it mean to the business? Who owns it? Who works with it? What's that rich semantic context around your data sets that makes them consumable and discoverable and ultimately becomes that that kind of foundation for democratizing access and use.
So the way we deliver all those capabilities is what we call the intelligent uh the Informatica intelligent data management cloud uh IDMC and IDMC is uh atscale platform and we process about 110 trillion transactions a month for our customers. So thousands of customers around the world on this platform and the way the platform works if you look at the diagram is we're based on this foundation of metadata and metadata is kind of a key theme you'll hear throughout the talk today and metadata becomes a critical enabler of both analytics and AI and that's at the core of our platform but on top of that we have a series of microervices that deliver different data management capabilities whether it's integration or quality and cleansing or API management uh all the way up to that governance and cataloging and master data management cap. capabilities and on top of this in the world of AI we have taken and created a layer of co-pilots and agents on top of all those services.
So as you go into the AI world you now and we announced these the last a few weeks ago at our customer conference at Informatica world you know we're going to have agents who do things like data quality or master data management to say clean clean is data for me right and be able to figure those things out using the power of generative AI. So the platform itself, it truly is cloudnative. It's offered as a service. Uh it's optimized for multicloud and hybrid environments.
We assume that data is going to live in different places and you can manage data processing, data management consistently and efficiently if you have a distributed heterogeneous environment. And because of that, we do have a security by design architecture meaning that you control where the data processing happens. It's all a SAS experience, but you actually can ensure that the data remains inside of your environments. we don't ever actually have to see it in our cloud technology.
We're dealing with your metadata and the processing occurs local to where the data is. And then uh we also have consumption based pricing. That's another big advantage when you think about all the different capabilities in this platform. Uh what when you look at mixing and matching those from independent software vendors, different pieces, you could be dealing with six or seven or eight different companies. Informatica lets you purchase a single unit of consumption for data management, an Informatica processing unit, and you can use it across any of these services.
So as you need data quality, as you need integration, it's very flexible and gives you that dynamic nature to address your problems quickly. It's also available on three, it's three cloud platforms where data bricks as well. So whether you're AWS, Azure, or Google, we're there and we're around the world available. And it really is designed, our goal is to be that complete end-to-end data management capability around the data bricks data intelligence platform to really address your most complex enterprise environments.
And when you think about the data intelligence platform, the data bricks data intelligence platform, uh we spend a lot of time with our product and engineering probably more so than any other technology partner data bicks has integrating and certifying our capabilities whether it's supporting unique catalog across data integration and synchronizing metadata pulling in deep lineage into our global lineage views with our data cataloging.
We have a rich set of integrations and we really do focus on trying to provide best-in-class best of breed and also being ahead of the game. You'll see some announcements a little later on today about how Informatica has participated a lot of the announcements you heard this morning uh in the keynote. Okay, so switching gears to AI. So today we're going to focus on what we call uh gen basically Informatica for generative AI.
So you want to leverage generative AI, you want to create AI based applications, your enterprise, you know, how do you do that? We also have a rich set of capabilities we call CLA which is our AI engine. We've actually had it in our cloud platform for more than seven or eight years at this point. uh originally focused a lot on machine learning and different techniques to accelerate the business of data management. So there's a whole host of things.
This is where our AI agents come in. Uh we have a GPT experience for actually interacting with data management from a prompt, but we're not going to go into that so much today. We're going to talk more about when you want to address your business cases, how does data management play a critical role. So, so this is a really interesting survey that was done quite recently, a CDO in CDO insight survey about, you know, where the state of Generative AI is.
And an astounding 92% of folks are, you know, either they're basically looking at their pilots that have gone to production. They're saying, "Wait a second. We're really concerned that we've skipped some important steps. We really didn't solve our our edge cases, but we rushed ourselves to production." And only 67% you know have or sorry 67% have been unable to transition though those ideas those new use cases from pilot to production and frequently data and data readiness is cited as one of the key impediments to getting there.
So you know 56% cite data reliability as a big challenge and in terms of the top issues that are preventing the pod production 43% ranked data issues as the main reason and so why is data such a challenge so we think about what does it take to deliver generative AI applications for enterprise scenarios one of the first challenges is that uh well I take a step back I think in today's world the way the public LLMs have grown and matured uh uh and also the complexity and difficulty of creating your own large language models or even fine-tuning.
Most customers are using public models, right, for that conversational experience. The problem with the public model is it does not understand your enterprise reality. And where is your enterprise reality? It's in your data, right? So being able to bring your data to the party, so to speak, is critically important. But also, you want to make sure you bring trusted data. And we found that master data for example customer 360 product 360 supplier 360 actually becomes hugely valuable data for grounding of generative AI conversations highly trusted data the next challenge becomes well these models don't understand the language of your business right it could be as simple as your user types in customer do they mean a company or an individual that's very specific to your business right or they type discount rate are you a mortgage company or a retailer very different context so being able to bring semantic context for example from governance layers to the to do prompt enrichment to do summarization incredibly valuable.
It also helps create explanability right and better understand the results that are coming back in the descriptions creates a whole degree of richness to your applications. The third point is is data quality and how that plays a role. And so one of the challenges you have when you have hundreds of data sources and you have agents now making decisions about where it goes for data. Well the not every data source is created equal, right?
Right? So having a a good practice around data quality and data quality metadata actually makes the agent smarter. They can tap into that data quality of scores and rules and make decisions about well I'm going to take the most authoritative data source and that's what I'm going to use for grounding. Again makes the quality of the experience that much better and that ensures access to the application. The fourth area is deployment right?
So there's a lot of tools out there to build agents to do agentic experiences but a lot of them are coding based. The problem with coding based is that it becomes very hard to maintain over time and it's very hard to evolve. So with Informatica and AJ my colleague here will be joining me is gonna kind of show talk about you know how we allow for no code definition of very complex agents and uh and gener application workflows right within our platform.
And last but not least is governance right so again this part of this is explainability and traceability. Another big piece though is ensuring that your data access policies equally apply in these types of applications. Just like when you're looking at reports or dashboards and limiting what users can look at, you need to make sure that when responses are being summarized and created that those are in compliance with your rules and the credentials of that user who's actually interacting with the application.
So we talk about well how does Informatica deliver those challenges? The first is helping you prepare that data environment right and where this is where we provide no code low code data pipelining data integration technology uh supports multiple modes ETL ELT change data capture we have many mo many modes of connectivity and because of all those different and we support about 300 different unique source system types for birectional connectivity uh but when you look at those 300 sources combined with all the modes of integration actually ends up being about 50,000 unique combinations of how you want to move data and get it ready for your application.
And this is support for both incremental and load incremental and initial loading of course on an ongoing basis. And I mentioned master data before again high quality master data. It truly is a tr the solid foundation for so many of these applications. And I mentioned a few domains earlier, but then foratica we have domains like finance 360, we have pro patient provider 360. So we have all these industry models that take all this a very complex set of data sources and rationalize it down to that clean foundation of data and that represents your core business entities.
And of course we are optimized within our pipelines and workflows to work with mosaic AI models if you want to you know take this the data that you're bringing in uh some of the data sources you want to vectorize and prepare them prepare them for semantic search. So we support very easily uh integrating that with the services from data bricks for generative AI. Now on the metadata side of things, this is where our cloud data governance and catalog comes in, right?
So the cloud data governance and catalog is a is a combination of two layers of catalog really. It has a technical view of the world and then it also has a business semantic view of the world and matches the two together. So that I mentioned the examples there bringing that semantic context and leveraging that through our APIs in your generative AI application makes those applications just so much more effective and efficient for end users.
And they mentioned also data dynamic data set selection. So when we run data quality, you can profile all your data with standard quality rules as well as a whole host of custom and industry pre-built rules. So you can run hundreds of different complex data quality checks to identify where there are issues in your data set. And that rolls up to an overall quality score that then can inform your agents and your gender of AI applications.
And the other mention also explainability and traceability, right? Again, being able to tap into governance and lineage. So when you present a result with data, you can actually have an experience that says, "Look, great. I saw the data. Can you show me where the data came from?" And because you're tied into this catalog experience, you can actually render that. Uh, and we do a great example. We don't have it here in today's session, but if you come by our booth, you can actually see we call our CLA GPT rendering inline lineage to kind of answer those types of prompts.
So those are those are kind of the key things here about the value of having a strong a good governance set of practices really accelerates the ability to deploy AI at enterprise scale. And last but not least is that no code framework I mentioned before. So this is kind of a visual of what it looks like. There's a whole pallet of controls whether you know for mosaic services, other services and it really is just drag and drop as well as we have a number of predefined recipes or templates for very common use cases in agent definition.
And AJ is going to walk you through much more granular view of how we actually recommend kind of our best practices or best our blueprint if you will for creating agents and agent-based experiences. Um, last thing I'll mention is that it's very easy once you create one of these workflows to publish it. But we also have full API management in the platform. So you can put an API in front of any of these complex process flows defining your your generative AI application.
And that API management actually lets you throttle and govern and rate limit which is really important where we're doing piloting controlling costs and that sort of thing. Um, it's a the huge priority for customers as well. So you can very quickly go from an idea and a concept and something that can be iterated very quickly on and get from pilot to a production environment quickly. So with that I'm going to ahead and bring AJ Gopali my colleague up.
So AJ is our director of solution architecture at Informatica for data bricks and he's going to talk about our agentic framework blueprint. Thanks. Thanks. Am I audible? So one of the things that when I was talking to customers right when they wanted to go with generative TVI application they said like this is what I was talking with the data manager over there he said like I said we want to take our data and make it a geni application but my biggest challenge is how do I limit the users to see what they can see because I don't want to give all my geni data and give it to everybody in organization the same level of access but I want same data most of the data to be repeatable that's one big challenge because today's technology we have solved it by adding role based access controls attribute based access controls and everything but this is a real life situation that I was talking to him and then he came back and said told me this it made us you know one of the advantages what shared across right not bringing their data making through the entire data journey with Informatica's intelligent data management cloud whether you are discovering the data or whether you are able to add quality and then making it governed making it trustable and high quality high fidelity data and making it easily accessible by your analyst nave it is all very easy to consumable for a consumerfriendly enterprise customers but now with generative AI the challenge is how do I use this data how do I make sure that this application data is easily transferable to my generative application in the enterprise.
So what we did is this is our blueprint for agentric framework to be able to utilize the data that you can take your agendi applications. We talk about three agents in this concept. The three main agents we talk about are the planner agent, the orchestrator agent and executor agent. The fundamental approach to this entire model is to use your genai applications have the trusted high quality data that's governed to know where is the data coming from.
Is this data the right data for putting into a generative application? Because sometimes you all have done this in the past. We have the data copied over to Excel sheets. I don't know how many of you do it. I do it. I copy into Excel sheets. Add my own insights into that data Excel sheet and then use the Excel sheet data to give it to sometimes I give it to and um smells my BS on the data and then he calls me out. But essentially what we do is we want to make sure that is the data right data.
What we are giving to your management is it the valid data because they are going to make decisions out of the data. Sometimes it's very scary that they are going to make decisions because your data if it is flooded with more assumptions it can go into wrong decision making. So that's one of the things that we use as agents. We have to be extremely careful in terms of providing all the data and applying it on generative TVI rather than you should know what data goes into a generative application who is accessing it why are they accessing it and what is the purpose for them with this in mind right so let's see what the planner agent does normally when you talk about NLP when you ask a question when somebody's asking a question in the organization what happens with the planner agent is the prompt engineering goes back in the back uh into the ll M itself it captures and validates the user intent.
It's not about just the saying of what I want but also needs to capture my sentiment. It needs to understand that hey is this person the right person asking for this data. So you need a kind of an identity management layer. You don't want to give your financial data to your uh sales guys. Definitely not. And at the same time you don't want your CFO to be restricted with not getting his data what he needs. So you have to make it make the entire thing in a robust way that your planning prompt should contain the conversation history.
What happened in the past question is obviously related with the intent of understanding the questions and also to understand what is the user input who is the user asking for it. What data are they looking for? Is it the valid data for them? Is it the high quality data or the low quality relevant current data or is it the high quality analyzed data? So there are a lot of things that we do in data bricks especially with go bronze layer, silver layer, golden layer what layer of data goes into your generative applications.
So the planning prompt understands the intent. It it decifers what the user is asking for and it provides an API management for filtering out any kind of D of service attacks and rolebased access controls and also it provides a usage governance you know you the easiest way to oh you know there's a genie who said like how do you spend billion dollars in a quick way spin up some servers and let it go or keep it running on any of the hypers hyperscaler cloud right so how do you limit the usages how do you Make sure that the models you use are not asking being asked the same question repeatedly multiple times and you're spending your unknown dollars over there.
So that's main idea about the entire planner agent. The agent does all these transactions. It keeps a planner prompt. The prompt contains what systems are onboarded, what data is available for the gender applications, what is the contains, who is the user, what identity management do they have, do they have the right access, everything goes in the planner agent. The next one is the orchestrator agent. Orchestrator agent is like the main central nervous system of this blueprint.
It knows understands the metadata information, the metadata what has been secured and using the governance catalog. Now it provides saying when you say I want to know what my order is in the supply chain management it knows order means it has to go into the uh sales system order system instead of not going into a billing system. So there are couple of questions. So it needs to be able to be intelligent enough to understand based on the user input to go approach which system it has to uh request the information for and it know it also need to plan based on the planner agent.
It needs to know what is the sequence of events that the planner based upon the user input is. The third agent we're going to talk about is the executor agent. These are the worker nodes. These nodes tackle the data. their their main job is to know like what is the filter condition, what is the user looking for, what is the who is the user, what is the role based access control or attribute based access control behind the scenes and how do I get this data for the user.
So the entire thing is brought over to say like the planner, executor, orchestrator and executor all work in co uh tandem to make sure that you get the best response for the question you ask. Now what we did right when I said this about when we put this framework a few months back and came back and said like I said show me this works. So what we said is okay we'll make it work we'll make you'll show it how it works because we want to not do a demo of it like just a demo but we want to show a real sample deployment reference architecture not just architecture but the entire demo itself.
So what we did is we put across a supply chain copilot a real life scenario where we have data coming from Salesforce for orders where we have an Oracle database coming with uh with the information from uh supply chain management under master data management providing parts and components and a transportation risk report sitting in data bricks. So multiple multiple data sets coming together forming a forming a complete use case scenario.
So think about this like a a supply chain analyst in a day if you go ask a supply chain analyst saying hey I want to know what order it is oh wait a minute let me go into Salesforce okay order spending get me the orders and then what they do next and then you'll say like hey what is the order which parts is the order contain oh let me wait I'll go to se different system and get the information and if you ask him more and more information they have to go into different different systems which normally takes them a lot of time from a productivity ity perspective.
Now what we did with generative is making it enterprise ready for your consumption for modernizing your analytics into a natural language processing unit. Don't believe me? Let's see a demo out of it. So the first question I'm going to ask is like a supply chain order analyst, right? So I'm going to ask like what is my pending orders right? So normally what is my pending orders or show me pending orders the way we do it.
So when I and I ask this question what's happening behind the scenes is I said like show me pending orders and the system knows that what I'm looking for my intent is to look for orders. Let me zoom this a little bit. So the system knows that what I'm looking for is orders and orders should have a status called pending. So the orchest the planner understands my intent, my question and it builds the plan and hands it off to orchestrator.
The planner does not know where the system is where the data is actually located. All that the planner knows is AJ is looking for the particular data that is relevant for his needs and the orchestrator knows exactly now it's filtered. So this is the front ending application. Think about our multic end architecture frameworks. We don't put our data in the front ends. We put our data in the back end with the MVC patterns.
Similarly to that pattern, now your data is not in the front end but it's available to the front end through an orchestrator. The agent now understands that question is saying okay now the user is asking for orders. Let me orchestrate it. Let me ask my worker nodes to get the data filter the data based upon the user persona and then sends it back again. So now if I dwell if I dwell deeper into this question. So I'll say show me details of the product ordered.
So when I'm looking for details of the product ordered, I didn't specify what product I I I am looking for. It went to a it's going to a completely different system right now. It knows from a previous conversation what is the product that's been ordered based upon the order that is in context. So the planner prompt knows what is in context. It understands what is the user's intent again. It understands that the user is looking for the product that's been ordered in the previous question and it executes it.
It goes from the planner to the orchestrator back again and then it orchestrates with okay I have to look into the details in the product from a master data management and it goes to master data management executes the SQL query behind the scenes and generates the data for the user. So all while all this is happening right this is all happening based upon the metadata information it is not not happening I'm not talking to the system directly but I'm talking to the metadata and the metadata transports the data back from the original repository into the front end so that's keeping the entire data secure that's one of the most important things so for example I want to let's say I want to see the supplier I'm going to click So when I ask about who is the primary supplier of this GPU again as I said earlier right so the context comes from various data elements I could have asked like what is the storage or give me the data for power supply I don't need to specify what is the product details are looking for but automatically the three laid name spaces brings back the data to show who is a supplier completely from a different system together right now.
So that is like think about it an analyst if you go and talk to them it would have taken at least one hour for them to fish for this information now they're able to do a NLP mechanisms with all the different data elements sitting in your organization. So that's one of the biggest things of generative AI that opened up organizations is to look for the untapped data knowledge without able to go and search and dig deeper and uh dive into the into the repositories.
But now it's all in your fingertips and this all being transported through the Informatica's data management cloud's metadata intelligence. How does it work? Right. Um, one thing that I did not tell you is when you looked at the entire query processing, I'm going to open up this query processing to show you when I asked this query, it went to first step is to go to the API management layer. It said like is this person asking it repeatedly the same question?
I'm asking relevant question. So if I ask go back to the same chat bot and then say like hey what is the capital of America? Think about it. Every single time when chat GPT came in our first I don't know many of you but I I'm guilty of it saying that the first question I'll ask is like what is the weather today and show me the I'm at the capital of Denmark and I was asking all these irrelevant questions but you know when something new is given to you you'll always want to play around to know its boundariness.
So people do that because it's human tendencies to do that. So when you ask not relevant questions, the system should be able to tell you, hey, it's not relevant to this particular chatbot. I don't want to waste my dollar going into an LLM to fetch what you don't want to know about. So that's one of the things that godridge that we put across over there in this uh in the in the in the agentic framework. So for example, if I say what is the capital of America, I'm being a human being here.
I'm not trying to be a a technical person. And I'm trying to be a a normal human being who will ask these questions on any any chat GPT interfaces. It'll also tell you saying that hey this query regarding what you asked for is not aligned with this functionalities. I'm not going to spend LLM capabilities to give you the information that you are not supposed to get. You can ask any question for that matter. Uh and then it'll tell you if it is not relevant to the systems that we onboarded.
It is going to tell you that it is not relevant. That is the most important thing uh in the in the that's what the API manager looks for and communicates it back to the planner saying is it the right question is the right context right relevance and then it goes into the makes a plan that planner provides for example which system is looking for information what is the filters what are the prompt selectors what are the different information coming from which supporting system provides this information everything gets built within this planner similarly if you look at The next question that we asked right um show me the uh who is a suppier supplier of this GPU very similar pattern it goes from the planner planner makes a plan orchestrator orchestrates does the routing of this entire algorithm executes to fetch the data and summarizes everything back into a natural language processing and brings it back to the front end and your front end could be anything it could be from uh teams copilots or it could be your genies or it could be your Amazon cues I don't know what you use are web apps.
You can use any of those. It can be integrated very easily and rapidly with this blueprint. Um in fact uh what we did is u we put across this entire thing with the metadata intelligence that I wanted to open up a planner that shows the execution plan from a metadata perspective. So if you see this metadata the metadata tells about like which system it contains where is the system located is it an Oracle system or is it going to be in a datab bricks environment and the metadata provides the entire knowledge that's been built to tell you the customer to tell the orchestrator where should I go for this information what should I get it from for the filter conditions so metadata is the starting point and then it goes to the actual reposit to fetch the information itself.
I can keep going and going uh on this once. Um I love this because I've been working on it for like last what like 20 months now. Uh so any questions definitely ask us. But most importantly right the most importantly what we did to make it easy for you is all of them is a is a quick no code no code. I did not write a code. I didn't have to write a single line of code on this one. That's the beauty of it. Everything is based upon as you saw the plan.
The planner works with what we call as a CI the cloud application integration fully based upon rest APIs and connectors with datab bricks with JDBC connectors with data bricks to so all the vector stores will fetch the data from uh from the individual data store repositories but the metadata is available through a vector store within the databicks environment for us to fetch this information. Uh that's the end of the demo but these all the deployed versions of it are available as a recipes or temporary awarded templates in our AI agent engineering uh in the services and you can quickly go to the IDMC framework.
When you have IDMC you can click on it while which framework which template you need deploy it and you make customize it based upon your organization needs and you can start utilizing it. That's that's exactly the advantage of what we made it to work with um with the data entire data management layer itself Thanks. come by our booth. We have a pretty big setup on the expo floor. There are a lot of cool prompts you can do.
I like playing around with it and see if I can break it all the time. It's a lot of fun. It's actually pretty really pretty really pretty really robust and uh it's a good example though of I hear from a lot of customers how do I do multissource rag chain effectively and it actually it's can be a challenging problem if you don't have the right metadata foundation so metadata is kind of key there okay um just kind of wrap things up and we'll definitely take some questions so start thinking of those if you have them um are the and uh actually these sessions already have they're they're recorded recorded okay good all right sorry the um these are two other sessions we did this uh this week here at uh data bricks data and AI summit Uh again one of them focuses on master data in a much more uh detailed in-depth view of what that all entails and then also we talk a lot about governance cataloging how do we integrate with unity catalog and what are the benefits of using our cloud data governance catalog together with unity and uh you can scan these QR codes recorded sessions um also if you want to learn more uh we do actually have a free service uh starter point for data integration data pipelines uh it's on the uh the partner connect we also have you can also just visit our website, learn more about all the great solutions we have here with data bricks and also our latest announcements um which actually I believe are coming out in a few minutes I think all right at market close so you'll see our latest innovations with uh with data bricks

Learn more about the Databricks Data and AI platform.