 And welcome back to theCUBE's live coverage here in the Lake House. We're on the floor of Databricks' event, Data Plus AI, CUBE coverage, I'm John Furrier. Rob Stretches here as well, wall-to-wall coverage for two days on the ground. Also, it's Snowflake in Vegas, Google event up in Seattle. We got all kinds of action going on. This is data week. And up here we got a great expert. We got Hema Raghavan, who's the vice president of engineering and co-founder of Kumo, expert in data retrieval, indexing, search, knowledge graphs, data. This is your wheelhouse. Absolutely. Well, first of all, you're part of the news. They announced Lake House apps, which you were mentioned in the keynote. Let's get into the news and we can talk about what it all means. What was announced? So, Kumo.ai is an AI platform company. We make AI as easy as SQL. And today we're a SaaS platform. It requires us to export your data into our SaaS platform and for us to be able to train the large models that we do. But that does require a security review on both sides. For us to have a very God rail system that we maintain on our SaaS platform. And likewise for our customers to be able to review and authorize that customer data actually leaves their ecosystem and comes in and is processed in the Kumo SaaS platform. With today's announcement, what that means is when we're working with customers whose data is in the Databricks ecosystem, in Delta tables, for example, the compute is actually pushed down to Databricks. So data never leaves Databricks. You get the goodness of Kumo without data ever leaving that ecosystem. Data is a big controversial topic when you talk about moving it. Expensive, moving compute to the data is a well-known discussed tactic. I'm even hearing about apps moving to the data. Workloads, not just compute, but workloads and compute. This is a big trend. And your pedigree in your background is modern application platforms. You worked at LinkedIn, you built that platform out. Platforms is the big discussion here. And Aoi is a platform guy. We love him for that. So are we here at theCUBE. The importance of a platform is a key to their strategy and open source, obviously a winning hand there. Why is that so important to enable these apps? Because platforms have to have an enabling aspect to them. But sometimes they may not be best to breed, but that fosters an ecosystem, some white space. Can you share your thoughts on reaction to that platform strategy and how that will play out for the ecosystem and for entrepreneurs like yourself? So let's take the example of Kumo itself, right? Kumo is a 30 person, one and a half year old startup. Our asset is that we run graph neural networks on relational data. So we're bringing the deep learning revolution to data warehouses. Now, a platform that actually allows Kumo to do what it does best, which is develop the algorithms that it has the best in-house team to develop and enable that for the world. Without having to think of security, governance and everything that comes with it is what a platform, like a true enabler like Databricks does in that ecosystem. It lets us focus on what we do best. It lets them focus on what they've learned to do best. And why is that important? And it's why does the ecosystem obviously developing? What is Databricks good at? If you had to kind of clarify what they're strong at that you like, that you think it's important, I'll see that doing the heavy lifting for the platform so I check. What else do they do that you like? So they definitely, I see them as, you know, definitely enabling the AI and analytics ecosystems, for example. So let's take an ecosystem, let's take the example of a company that, you know, is on Databricks and has Kumo as well on it. Kumo is great for if you want to build recommender systems, for example. Things that me and my co-founders did many times over in past lives, right? And so if you want to bring your data, you know, your data from your website, the tracking information, your Postgres tables, your Delta tables, all of those tables back into the lake house. All the old stuff. The old Avenue stuff into the lake house. Time series included. Exactly. So you bring all of that into the lake house and then you want to build your recommender system. So Kumo would plug in right there. And then the output of Kumo can go into other applications that are powered by Databricks, right? For example, Databricks allows building a service that can actually, you know, serve these models and these recommendations. So it fits very nicely into the workflow and everybody doesn't have to become an expert in everything, Databricks becomes the enabler and companies like Kumo become, you know, the expertise providers. What's interesting too is that it allows you to take what you're great at. You said recommendation systems, you built that in other platforms. We know those big platforms out there, you know that out there. But when you hear the keynote today, JP Morgan, Chase, executive, he talked about personalization. Personalization is pretty much in kind of what you do, right? And this is going to be part of every app. Exactly. Because personalization takes advantage of data domain specific, domain specific data. Yes. But has to work with horizontally scalable data sets or data products. So we're seeing this role of an emerging role that we call the data developer. Yeah. Kind of like you guys. You're kind of like a data developer. Absolutely, or algorithm developer or, yeah. Tell us about Kumo. How does it work? So you're building algorithms, you're building systems for data to move around, make things accessible, graph databases, neural networks, which by the way, sounds like it's really perfect for the AI revolution. Absolutely. Share a little bit about Kumo. So as I mentioned earlier, Kumo tries to make AI on data warehouses as easy as it is to do SQL. Today the AI revolution touches unstructured data, that is data that's text, that's images. But a lot of your enterprise data is actually in tables. Tables that are relational. Think of any application. You have your user data, you have your product information, you have your transactions, you have your clicks, you have all of that sitting in individual tables and across the organization. Our LLMs aren't working on that. So how do we bring that deep learning revolution to enterprise data, the data on warehouses? What Kumo does is we make the process a lot easier. So let's double click a bit on that. Doing AI on data warehouses is a lot like building question answering systems in 2003. That's what I did my PhD in. And it would take us a year to build a system. We would have to understand each word, we would have to understand part of the speeches around. The taxonomy, the word combination, the linguistics. Exactly. But we don't do any of that anymore. The LLMs, the hidden layers, figure all of that out for you, right? So how do we bring that revolution where we just give columns, we give tables, we give the metadata and a deep learning algorithm actually figures out what's the representation. And so Kumo brings that, we turn the relational data warehouse into a graph representation and we run graph neural networks under the hood using AutoML. Yeah, Kumo, I think you're a great example, first of all, great entrepreneur, congratulations on that. That's hard to begin with to do that. But you highlighted two things that I think I want to call out. One is that you're a data product. You're building a product. Yeah. That's basically a data product. It's a data product, yeah. What do you call it as a product? And you have developers and you work with that. But you mentioned all the old school days, those were heavy lifting, hard things to set up. Yes. In order to get to the start line. And think of the people you had to hire to get that done because you needed, when I was at LinkedIn to build a recommender system, I needed to hire engineers who were unicorns, who could write workflows, who understood how. And they get 10 job offers a day from other companies. Exactly. And so now you have that opportunity. That kind of reminds me of how the cloud started. Remember when I did my first startup when the cloud was there, I didn't have to buy a box. Yes. I could just put it in the cloud with simple primitives, EC2, S3 and Q-ing. Yeah. All there. That's all I needed. Yes. I got going. I'm a SaaS company. Yes. Dropbox, Airbnb. Yes. Now data with LLMs gives that same ease of deployment. Yeah. Low cost entry. Yeah. Or as Ali says democratization. Exactly. I mean I call it more innovation less documents initially for entrepreneurs. But this low cost entry to get started. Yeah. Allows for more creativity. Yeah. How many companies have died because they did too much setup, couldn't get the cash. Yes. Or energy people to build. Exactly. Now we're in a builder mode. Yeah. I think you hit the nail on the head. So today to build AI on warehouse data is like that. It's like building an on-prem data center. The analogy that you just used. Shoot me in the head. Punch me in the face. Or. No one wants that. Nobody wants to do that by the way. Absolutely. So that's what we aim to make easy. We want to make AI. Anybody should be able to do AI. And we want to be able to let the data scientists focus on what they do best. They know where in the business you use the models. They know what data is important to build their models. They don't have to do all the grunt work to put together pipelines. It's really, I do agree with Ali. I've been saying this on theCUBE for over a year. It has eight months hardcore. It really is a great time. And I've never been more excited even at my age. I'm old, I mean. But I wish I was 25 again. I say that all the time on theCUBE. I mean, final question is, do you think these young kids actually know how good they got it? I mean, think about it, if you're like 25. I'm 25? I wish I was 25, no. If you're a young kid, we used to walk in the bare feet in the snow. That's true. Building out databases and standing up all this infrastructure. Absolutely. I mean, it's the perfect time for developers right now. I mean, open source is booming. Think of writing code. I mean, think of the time when, I learned to assembly when I was in undergraduate, so. Register code, core dump, debugging. I mean, come on. I think you can go through an entire CS program without knowing anything about assembly language or operating systems today. That's how the AI revolution's going to be for. Well, hey, but you're great conversation and I'm super excited for your venture. Congratulations on being in the announcement. I'll give you the final word. Put a plug in for your company. What are you guys looking to do? What are you guys hiring? What are you guys trying to do this year? What's your objectives? Put a plug in for the company. Absolutely. So, Kumo aims to be what the one place that all data scientists should choose to build their AI applications and data warehouses. And so we're growing. We'd love for you to apply. Please try to us at jobs at kumo.ai and please write to us at hello.atkumo.ai if you want to try out the product. Thank you. Okay, this is theCUBE coverage here at the Lake House hanging out with the Databricks folks here bringing all the coverage wall to wall. I'm John Furrier, Rob Stretcher, my co-host analyst with theCUBE Collective here. I'm getting all the data sharing that with you. Of course, we love AI. We got a lot of data and hope you're enjoying it. More coverage after this short break. Thank you.