Sponsored by: Tenarai | AI-Accelerated Real Time Supply Chain Operations on Databricks

Summary

Tenarai (formerly operating as Tredence in earlier contexts) works as a forward-deployed engineering company in 20 countries with 5,000 employees, applying real-time streaming and AI solutions—initially developed with technology partners like Meta—to supply chain operations for global enterprises.
The company's supply chain solution on the Databricks Data and AI platform uses Structured Streaming, Delta Lake, Unity Catalog, Agent Bricks, and AI/BI Genie to address demand volatility, inventory misalignment, shipment risks, and decision latency caused by fragmented batch-era architectures.
Tenarai reports outcomes including approximately 20-plus percent faster operations, approximately 15 million dollars or more in cost savings, and approximately 60 percent reduction in manual effort, with an 'Archer' agentic team that automates migration tasks including schema design and SQL conversion.

Chapters

FAQs

What supply chain problems does this solution on Databricks address?

This video focuses on four core supply chain challenges: demand volatility that makes forecasting unreliable, inventory misalignment between what is stocked and what is needed, shipment risk visibility gaps, and decision latency caused by fragmented systems and batch-era architectures. The solution uses real-time streaming and AI on the Databricks Data and AI platform to connect inventory, transport, and predictive models in a unified pipeline.

What is the Archer agentic team and what does it do?

Archer is Tenarai's agentic system that automates supply chain platform migration tasks, including schema design, SQL conversion, and workload modernization beyond simple syntax changes. This video presents Archer as a key accelerator for enterprises moving from legacy batch architectures to modern real-time pipelines on Databricks.

What Databricks technologies power Tenarai's supply chain architecture?

The architecture described in this video combines Structured Streaming for real-time data ingestion, Delta Lake for reliable storage, Unity Catalog for governance, Agent Bricks for agentic workflows, and AI/BI Genie for natural-language analytics. Together these components create an end-to-end pipeline from factory floor data to AI-driven decision support.

What business outcomes did Tenarai achieve with this architecture?

According to this video, the solution delivered approximately 20-plus percent faster operations, approximately 15 million dollars or more in cost savings, and approximately 60 percent reduction in manual effort for supply chain teams. These results are attributed to replacing fragmented batch pipelines with a real-time, AI-accelerated architecture on the Databricks Data and AI platform.

Full transcript

[00:07] Hey everyone. Thank Thanks for being here. Uh I'm Sriraj Srinivasan. Mod couldn't make it today. Uh We're from Tredence. Um so today's talk is going to focus on supply chain and how we bought in Databricks from a edge
[00:19] and cloud computing perspective to actually accelerate real-time operations. So I'm just going to start with a brief introduction of who Tredence is for a lot of folks who don't know about us and give us give
[00:33] a view of the customers that we work with and also the type of work that we do in Databricks. Then get into what what's the type of problems that we're working in supply chain. Different customers, different types of use cases,
[00:45] etc. Few demos in between and hopefully they work. And then finally just a quick recommendation of how we see Databricks being embedded in the supply chain universe.
[00:59] Right. So Tredence has been around for about 30 years, I would say. I mean, they've rebranded themselves several times. Today what we are is a company that is fully AI focused. We work with several partners. Databricks is one of our strategic partners. And we we
[01:12] operate in a forward-deployed engineering model. So for we are located in about 20 countries, 5,000 employees. And we work with this large set of customers in which we our
[01:25] journey started with a lot of work with technology. And especially some of the real-time streaming problems that we're solving today, you'll see we learned from working with Meta in the reality glasses, which we took to other industry verticals to the right. Right? And a lot
[01:39] of the work that is happening today within these technology vertical, Databricks is heavily embedded. So it makes it very easy to learn things and then move it towards the right. So this just gives a view of how we
[01:51] actually operate as a organization. We are forward-deployed engineers. We come in, build, deploy, then hand it off to customers and move on. And we are working operating in four different areas and like we work end-to-end. So,
[02:03] what we saw yesterday in Databricks the keynote where Ali talked about how they're moving to the apps. We Tenrai started from the apps layer and moved downwards. So, they moved from the apps to the machine learning to the data,
[02:16] then eventually to the core capability layers. And we partner with several large uh platform vendors. One is Databricks, the other one is GCP, Anthropic. So,
[02:28] depending upon the use case, we use the partners and actually run with it. So, we learn a lot through the process. That's why the FD model works. Right. So, I'm just going to start with the supply chain world as this talk is focused on that. So, I come from a
[02:41] manufacturing background. I got a PhD which focused a lot on manufacturing in oil and gas. And I used to build multi-physics models which all the way went from quantum physics to AI. And this was like 15 years ago and and
[02:55] it was so difficult to get high-performance computing even in the US that we used to actually figure out ways to run it on our inbuilt clusters. Today, all those problems have been solved thanks to Nvidia as you're all aware. And then we've got all kinds of
[03:08] AI models running. So, some of the key challenges we see with the customer, the first one is the geopolitical volatility, right? So, you are seeing what has happened in the war between US and Iran and the strait got closed, the oil prices went up. And this impacts all
[03:22] the industries, just not manufacturing and supply chain. If you're in retail, CPG, and I worked across industry verticals. So, we see that impact. And what we Well, while talking to customers and you saw the customer list that I showed, someone like FedEx is very
[03:34] impacted by what is really happening because of the geopolitical volatility that you're seeing. So, so does Petronas. So, does someone like a Walmart. So, all these customers we talk about the same thing. And then the
[03:46] climate disruptions that has been continuously happening which is all uncontrollables. These to to act mitigate these types of problems, what the shift we are seeing heavily within the industry ecosystem is moving towards
[04:00] real-time operations. This has been happening over time. And platforms do exist. It's not the technology issue, but more from the standpoint of a lot of these consumers and customers not having data together, right? So, the batch what we
[04:14] are seeing is the batch data era is over. Completely everybody wants everything real-time with an app on your phone. And the data needs to stream at millisecond latency. And it needs to be generating insights. It's no more visibility, no dashboards, nothing. They
[04:26] pretty much if you work with a Walmart, what Walmart is doing all of you are seeing is you place an order, they fulfill regardless of all these disruptions, the fulfillment is happening in 2 hours. And that is happening because of the back-end systems running on things like data
[04:39] bricks, Google, etc., right? And the other ones the Agent UKI. What Agent UKI has bridged the gap is between an SME say in supply chain to somebody who is like an consumer. So, that bridge has
[04:52] happened really well. And Agent UKI and that's what the genie is there what they talked about is actually doing a really great job. It's able to take what is uh simple semantic language that people are
[05:04] using, connect it backwards to data systems which could be inventory, transport, or any type of ERP systems, and then pull it out. That's the 50 plus sources, right? So, we've done this with so many different organizations, and
[05:16] some of them have built their own databases, which is over years SMEs have disappeared. So, a lot of things that is actually coming together because of uh uh the data being available, the real-time, and the AI.
[05:29] So, why is it really big for a lot of people uh is that the supply chain market itself is growing uh at a very rapid pace, right? It's a huge market about 10 billion. 94% of the
[05:41] companies are playing in that space. And then what are the key things that it's impacting is what one needs to look at. It's not to do with the impact on saying, "Well, my pipeline is broken." It's more towards to do with your P&L, right? Profit and loss statement. You
[05:54] look at 20 to 30% inventory reduction, that's a cash flow improvement. You're saying profitability is improved as they get more and more real time. And this is validated by a lot of our customers. And that's how we go into the customers
[06:06] itself. We are not talking about fixing a pipeline, instead we are saying we'll sign up an outcome to improve your inventory problem. We'll sign up for an outcome to reduce your or improve your profitability. That's what is today what
[06:18] AI is able to make us do. And And the last bit is very interesting. So, I mean, a lot of you who work with robotics, if you go into any of the warehouses and manufacturing
[06:30] sites today, it's almost 50% robotics, right? Because of the edge systems that Nvidia has built, how it communicates through the cloud system seamlessly on real time. Most of the labor shortages are being addressed because of that. And
[06:42] especially Amazon warehouses are 90% robotic, right? And you get into Walmart, they're about 60%. So, this is like a big shift. By 2030, a lot of them ambitiously are thinking of 2028 that they want to be 15% autonomous, right?
[07:01] So, what's really stopping people from doing this? That's the next big question, right? So, most of it is legacy. So, you talk a lot about people think every legacy problem has been solved, not really. There's millions and millions of records that has been built over like tens of
[07:13] years and nobody has migrated them. So, there is a big push. If you go to any chief supply chain officer or any chief manufacturing officer and talk to them, first thing they'll tell you is fix my data.
[07:25] Forget everything else, right? All this genie is good as an entry point. You can give me insights, but the back end data systems don't talk with each other for various reasons. So, come and fix it. So, that's where we've seen a lot with Databricks at least from the lakehouse
[07:37] architecture, how we can bring everything together in different zones, right? In the the the landing zone plus the curation zones and then make it consumption ready in like format free. Yesterday it was very interesting to see what they talked about like iceberg and
[07:50] Delta is no more. You can seamlessly interface between them using agents. That is a big shift in how it helps out quite a bit of what is happening out there. Then the talent gap. So a lot of folks that are there in supply chain and
[08:02] manufacturing hardly use AI. So companies want to invest the money, you spend like 30 40 million migrating data. But then there's no talent to actually manage it. And so there's a big gap and that's where this forward
[08:14] deployed model that Palantir and people before them invented is picking up much more because you want somebody who can look at the business, also look at the problem engineering problem, solve both of them at the same time. And then the value risk. Today every every engagement
[08:28] that you start, be it a consulting player like us or a products company or platform, every everything is hedged on the value. You want to go in and hedge on the value. The customer wants to actually share the risk with you. So it's all
[08:41] about hedging and we can accelerate it using AI which is what we are seeing from what you see on the right side that people are trying to do differently to address these types of issues.
[08:56] So what are the key supply chain problems, right? And I just gave a few brief introduction to some of them. The first one is the demand volatility. This has been the problem for the last, I don't know, 30 40 years because still people cannot forecast accurately. It's kind of
[09:09] crazy because you don't know what is happening real time across the world. So depending upon your make move sell value chains, what will happen is if you cannot see the data, not only your captive data of the transactional
[09:21] systems, but also the third party data. Like if a truck is moving somewhere in different parts of the US or a ship is moving through different parts of the world, people still don't get real time data. The streams are not real time. So that is the first problem. The demand
[09:34] volatility, people don't know how to address it. They start losing money. Second thing is tied to that because of the fragmented systems. The systems don't talk with each other. Wait, ERP, WMS, the warehouse management systems, the transport management systems, all of
[09:47] this, right? So, this is like a big problem that has persisted and and people are trying to fix it. The inventory misalignment is operating cash flow problem. You don't have inventory properly aligned, you're not going to fulfill an order. And we are seeing this
[10:00] with customers like Walmart's where that is the number one problem that we continuously address because that's tied to how on time in full they actually deliver the goods to your house, and also how they manage their cash flow within warehouses. They don't want to stock up a product like a
[10:13] chips or something else in a particular warehouse when there's no demand to it. So, number one and number three are tied, but the biggest problem for both are the fragmented systems, right? So, you have to address the fragmented systems problem. And finally, the
[10:25] shipment risk. It all completes the value chain. That's about it, right? So, And interestingly, the last one when we work with like logistics providers like who's a customer of ours, for them the shipment risk is a bigger problem
[10:38] because on the floor when the order comes in, they are not able to see how to distribute it to different people. So, whether the trucks are ready, the ships are ready, etc., right? So, there's an entire value chain problem that has emerged purely because of uh
[10:50] the fragmented systems. So, some of things that we've done and I'm going to show you some use cases that we've implemented, but before that, the approach that we've taken is how do we take So, look at the systems at the edge, right? That is the number one
[11:02] thing that is a problem for everybody. They don't get edge visibility within a warehouse what is happening, whether somebody shows up to work or the machine is running, machine is broken. Then, how do we communicate this information all
[11:15] the way up to the cloud systems? That is the physical edge. And a lot of these problems today are being solved by the Nvidia boxes. If you're working in the space, you should be very familiar with it. And then the streaming injection, right? The event hubs that are happening. Each event that actually
[11:28] takes place within a particular warehouse, which could be your equipment is broken down. Somebody has not shown up to work. A conveyor belt is not working. All that stuff is actually being handled
[11:40] by the streaming ingestion in real time, right? And there are different solutions. Kafka has been widely used, but you can take what we have from Databricks or any other like Dataflow from Google, anything else.
[11:52] So, the lakehouse has become universal now. If you are aware, Delta Delta tables and Iceberg are becoming the global like little industry standard formats. Everybody is putting it there so that anybody can access it and then
[12:04] they're acquiring it through the intelligence layers like using Genie, which is backed by Agent IQ, right? It's as simple as that. But now all of this needs to move between the edge and the cloud. So, that's where your serving
[12:16] decisions and the latency is determined by your compute systems, right? So, where how how do you compute? And then how do you choose your compute is super critical because one is that okay, you're paying a lot of cost for
[12:28] the tokens. The second bit is also computing the data, you're going to lose a lot of money. And a lot of the chief data analytics officers, that's where their problem is. Okay, you you build this cool solution you're showing us, but how
[12:40] much am I going to pay for this and what's the ROI, right? So, we're seeing both sides of the story, so it becomes very critical for us to pick a problem on the value. Right? If it's a high value use case, then people are willing to bet on such type of systems and I've
[12:52] been part of such engagements actually scaling it up in large manufacturing and supply chain for retailers, uh consumer product companies as well as for oil and gas companies, right? So, they see a larger long-term value, so they put the
[13:04] money into this. So, the outcomes that generally you see on the right side that we've seen actually works for most of the people is uh I mean, you're improving the supply chain times. Second one is on the cost to
[13:16] closing some of these programs. And Agent IQ AI, I'll show you some of the accelerators that we have actually help accelerate this whole thing. And consequently reduce the manual effort, uh the labor required. And and the last
[13:28] bit is how many source systems can you integrate? It breaks it slows down the fragmented uh systems problem that I was talking about.
[13:40] So, this is like a unified architecture that we've built, uh data and AI. Uh so, this is what I was talking about how you go from the physical location and all the way to the end user in one stream which is in seconds or milliseconds, right? And all the edge
[13:53] that we play today, at least we partner with startups that are working in Nvidia, because all the edge boxes in all the manufacturing floors today, either it's a Dell box or a Nvidia box, which is actually built on
[14:05] open source to stream data. It picks the data real time and is able to actually uh push it forward, right? And some cases it computes the edge inference itself, which means you can run compute on the edge. You want to make a decision
[14:17] on the floor, then it's able to actually do it on the floor. Uh and when I call floor, it means a warehouse or a factory or even store apps, right? You go into Walmart, you'll be surprised. You go in the morning and you go in the afternoon, the shelf,
[14:29] whatever the products they have on the shelf will change. And that is decis- that is a decision that is taken by the vision guided systems that actually sits around you. And then we use this I mean, I don't want to go through this, but what you
[14:41] see is the central cloud as we call it is all embedded on Databricks on any type of hyperscaler. And we work with customers with different hyperscalers, but Databricks is one of the universal compute platforms that they're using to actually drive the data and AI.
[14:54] And the serving layers, you can use any type of systems, right? So, I'm just going to jump to a couple of use cases demos. So, some of the data this is the way we
[15:06] actually look at how we go to our customers. So, I'm just playing it and it's no different from most people that would have done it, but when we look at different verticals, we these are the class of problems that we're seeing repeating, right? And many of you would
[15:18] see something else different. I'm sure, but so, and what we're seeing with the supply chain side, which I call analytics, is a lot of more physical AI coming in, where people are looking at the edge and the
[15:30] the cloud, but also physical AI is becoming critical to actually unlock the value. If you're not able to show anything in the floor, at the real environment, there's no value to it. And then, you can either work from a data
[15:43] monetization problem and move upwards and solve a physical supply chain or like a sales marketing problem, or the other way around, right? So, that's that's like a cycle that we're seeing emerge. And the moat that has made this happen is because of the
[15:56] I mean, I put it as a AIBI Genie Agent Bricks, but you take any other any stack, it's the Agentic AI that is creating that moat very easily. And sooner or later, this whole thing could be autonomous. You could be actually modernizing your data
[16:09] as as at the same speed as probably you're generating insights. Except that you need to allocate a lot of compute and figure out your compute systems, but I think that is getting closer and closer.
[16:24] So, from our standpoint, this is how we went up went about actually going in with Databricks. And we said, "Okay, this is how we're going to make ourselves specialized." So, we are specialized across the board in the different areas that I talked about. Physical AI is something that we've
[16:36] actually started exploring with some customers, and we've built it for with other stacks, but we are starting to explore with Databricks and get forward, right? So, so I'm going to start with a physical AI use case, which basically we
[16:49] built for a large steel manufacturing company. One of the largest in the world, and how we got in and I'm just throwing some of these things because for you all to think about it. Like, if you're talking about a customer profile, where would you take a physical AI use
[17:02] case? It's basically for manufacturing, CPG, retail, logistics, because it's a no-brainer. You have the vision going I mean, the camera everywhere, and then you're pretty much detecting three problems, there. Safety, productivity, and quality
[17:15] improvements, right? This is standard. So, if you've worked in manufacturing, these are three things that everybody wants to look at. And if you look at your P&L, every chief, I mean, CFO would say, "Hey, improve these three and your
[17:28] profit margin is going to improve." So, we we mixed Databricks with open source YOLO models, right? On the edge, we ran it on Edge Box and H100s, which is Nvidia, which you can just buy off
[17:40] the shelf. We implemented that and we connected it to like real-time equipments, which and I'll show you some of those. And what we calculated was simple, which is OEE. So, OEE is a metric that is used universally by every
[17:53] company and we just focused on saying that we'll improve your OEE by 5% by just using vision guided systems. And that's pretty much how we pitched and we continue pitching to different customers, but we implemented this for
[18:05] one of them and I'm just going to show you a demo video around. And I'll just walk you through the demo as well once it's done. It's pretty fast. So, this is like a steel manufacturing plant, what you're seeing, and then it's
[18:19] targeting three things and I know there's a lot of activity going on. The first thing it's looking at is it's checking if the operators are wearing a cap, safety cap. The second bit it's looking at the steel is getting dipped in some kind of
[18:32] a coating and it's looking at whether the dipping times are accurate because that determines how much coating goes on it. And the third bit is it's making sure from the data coming streaming in the back end that whether all the
[18:44] equipments are available at the same time. There should be no breakdowns because the throughput will get impacted. So, what we're computing in real-time stream is the OEE, as I told you. And each plant, depending upon the tenure of the plant and how much the
[18:57] equipment is older, the operators available, we are able to determine what the OEE would be. So, our sign-up to even customers is like, "Here's the OEE that you want you have today from our discovery and what would you want to
[19:09] achieve?" And based on that, we determine whether such type of a a solution actually works, right? And this the back end of this is all run on Databricks, which is running it the stream the data streams. The edge
[19:22] streams are coming from Nvidia. The backend data is all Databricks sitting on some hyperscale of the architecture that I showed. Right? And this customer we built it on Azure Databricks. This could be done in anything else. It's not a big deal. Yeah.
[19:34] So, based on that what we looked at we solved the edge problem and then when we talked to some of the customers including the same customer, right? What really happened came to our understanding was okay, we have tons of
[19:46] ERP systems. Now you're talking edge we gave you a particular data set. Fantastic. Okay, now you've solved the problem connecting all PLCs and SCADAs of the world. But the real IT data sits in like zillion ERPs. And every time the
[20:00] new CIO shows up they add an ERP I think. So, there was some 50 ERPs. So, now how do you really bring this together? If you want your transaction data and your edge data to come together, then you have to modernize the
[20:12] transaction data and that's where this journey of migrating to Databricks started, right? And because the choice was made purely that Databricks is a great compute system, right? Which can match in latency with the edge. That's
[20:24] where the choice was made as Databricks, right? Um and then same thing. What we we spoke about it and then the business problem was very clear to them because we started with the business problem on real-time visibility and
[20:36] decision intelligence. The the value case was already built. Now it was all determining how much workloads and from which legacy source systems we needed to move. And in this case we start we started the journey with for for like a customer like big
[20:50] logistics major, we started with like something like SAS, Synapse. We started pulling out data from like Teradata because these are all very low latency systems. And then we started applying agentic AI and I'll
[21:02] show a demo of how we built an accelerator completely on agentic AI including the team was built with some 12 agents. There was one product architect and 12 agents and we built that whole accelerator in a span of 3 to
[21:14] 4 months by source system and we started putting it into the customer environments through through an FDE model. FDE would get in and actually start building it and those with outcomes we signed up. So when we got in we said, "Okay, if you're not
[21:27] you're not going to give you 30% faster and 20% cost savings, then we'll pay for it." That's pretty much what we did. And I hope this demo runs. So, let's see.
[21:40] >> Archer is an agentic workbench for data modernization. It scans your legacy environment and maps every dependency before you move anything. It sequences your migration into structured waves
[21:52] based on actual code complexity, not estimates. It converts workloads to cloud-native targets and validates the output automatically. Okay, let's get started by logging into the Archer
[22:04] platform. Now that we have logged in, let's start by meeting the Archer agentic team. Archer is a truly agentic platform where specialized AI agents coordinate and
[22:17] work together across the entire data migration life cycle. Robin is the orchestrator. Robin coordinates Archer platform's specialist agents, plans multi-step migration workflows, and
[22:29] guides users through their data warehouse modernization from discovery to validation. Marcus assesses workload complexity and turns assessment snapshots into dependency-ordered migration waves.
[22:42] Marcus clarifies what should move first, what must wait, how much effort each wave will take, and where the delivery risk sit. Ash is our Databricks target designer.
[22:54] Ash designs Delta Lake schemas, optimizes Spark SQL, and maps source patterns onto Databricks native architectures: Unity Catalog, medallion layers, liquid clustering. Mira is our
[23:07] BigQuery target designer. Mira designs BigQuery schemas, optimizes for slot efficiency and cost, and maps source patterns onto BigQuery native features: nested/repeated fields, materialized
[23:21] views, partitioning. Fletcher is the code converter. Fletcher transpiles SQL across dialects, rewrites constructs SQL glot can't handle, converts stored procedures and BTEQ
[23:34] scripts, and annotates every output with confidence and divergence notes. Snow is our validation engineer. Snow confirms migration correctness through reconciliation queries, multi-dimensional comparisons, and test
[23:47] plans, the go/no-go signal before production cutover. Later on, we will go over the agentic workbench, where you can interact with these agents in a conversational interface. Start by creating the
[24:00] migration program. Enter the program name and description, select the source and destination platforms, and add any relevant program context. For this example, we'll choose the SQL Server to Databricks PySpark SQL
[24:14] migration program. Next, we'll create a workload. Enter the workload name and scope, then add the relevant source code, SQL scripts, and schema files associated with the workload. You can upload a zip file,
[24:28] pull from a cloud bucket, or import from a GitHub repository. For this example, we'll select the order-to-cash pipeline migration workload. Next, review your portfolio.
[24:41] Upload your source scripts and decide whether to include or exclude certain workload scripts. Automatically synchronizes updates, indexes your code, and provides a centralized workspace where you can
[24:53] review scripts and manage the workload scope. It evaluates every SQL script in a workload, providing detailed insights into both code and migration complexity. Each source script is thoroughly
[25:06] analyzed across a spectrum from trivial to critical, considering factors such as scale, complexity, structure, and dependencies. Leveraging its proprietary and soon-to-be patented six-dimensional
[25:19] complexity model, Archer delivers precise assessments of code and migration complexity. Archer platform provides enterprise-scale data lineage and impact analysis across SQL, ETL, and shell
[25:33] scripts. By mapping dependencies across thousands of objects, it helps teams understand relationships before migration, reducing risk and preventing critical dependencies from being missed. Archer
[25:45] will provide a detailed assessment report for your selected workloads that can be downloaded in Excel format. This report provides object-level complexity scores for offline review and
[25:57] stakeholder collaboration. Users can examine workload complexity, verify the portfolio inventory, review dependency summaries, check function usage, analyze referenced objects and cross-references,
[26:10] and consult the glossary. Archer's agentic workbench tracks your migration across the assessment, planning, converter, and validate phases. Using conversational guidance, it generates an automated wave plan that
[26:23] outlines the cutover sequence based on workload dependent team. That's not faster guessing. That's removing guesswork from the process entirely. Next, Robin will initiate the conversion
[26:35] process by instructing Ash to create a target schema definition by looking at the approved wave plan. Ash looks at the latest documentation and best practices to define the target schema.
[26:47] Once the target schema is finalized, the converter engine takes over. It ingests legacy SQL across six or more source dialects and automatically transforms it into optimized cloud-native code. This
[27:01] goes beyond simple syntax conversion. Workloads are restructured and modernized to improve scalability, performance, and maintainability. Engineers have access to side-by-side comparisons of the source and target
[27:14] code, enabling them to examine every transformation and decision in detail. No changes are deployed without approval, ensuring full control throughout the migration. Finally, the
[27:26] validator engine ensures output accuracy by running automated checks against your target schema, verifying data integrity and logic before go live. The process performs automated checks to
[27:38] confirm the target environment is prepared and that data is accurately reconciled. The platform reports conversion failures and provides guidance for re-running validations, helping identify issues
[27:50] such as SQL functions not parsing correctly in the target dialect. For example, 337 checks passed, four failed, and six were skipped. The validator engine covers schema checks, key metrics, and row-level
[28:03] validations, ensuring the target environment is fully prepared and data is accurately reconciled before migration. >> Good. Thanks everyone for listening to it patiently, but what
[28:16] we've done is we've actually taken this, shown it to Lake Bridge. Right? We ran a demo with Lake Bridge and they were they have a roadmap and this aligns with that roadmap, right? That's what surprised us, but I think this is open
[28:29] for everybody to think about that everybody is going in this direction. So, it's and we've taken it to customers, dropped it in their environments, and we are seeing the migrate-modernize timelines drop significantly. I mean, so it's real. In
[28:42] a way, it's scary that this thing can go so fast and we are talking about multiple source systems. SQL is what the demo run, but we're hitting all the source systems and the LLMs are doing a phenomenal job. Just just to call it out, right?
[28:55] Yeah, and based on this what we did, it is pretty natural for us. We did that for the more complex type of workload. I'm not going to run another demo and bore all of you, but >>
[29:08] there's so much of legacy back end stuff that we pretty And if you come to our booth, we are actually running these demos there. We have our solution engineers doing it, but what we did was we just understood that agents can do really well in the insights intelligence
[29:20] layer, and that's what yesterday was disclosed even by Databricks, but we've been actually testing it with customers and scaled it. So we got recognized by Databricks, as I showed, for this, where we bought completely agentic workflows on like using agent bricks. We actually
[29:34] move a lot of Power BI. You talk about Cognos. You talk about ThoughtSpot. All that stuff is starting to move towards Databricks ecosystem. So and this is
[29:46] fairly doable uh in a very short turnaround. So I'm just going to pause. I just have one last slide and that's about it. Um
[29:58] So our recommendations, if you're working with the world of supply chain because this talk was focused on it, is actually build the business road map, right? That is number one. I think most of you working would know this. It's not a no-brainer. But I think the timelines
[30:10] to go from the building the road map to actually delivering an MVP and scale is now gone down by 50% easily. So I think that that's the big learning for us. Integrating agentic AI at every level,
[30:22] leveraging the architecture that I showed, actually works at scale, and we've done this for many customers. Yeah. Thank you. >>

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.