 Hello, welcome back to the cubes coverage here at data bricks data plus AI summit I'm John Furrier host of the cube we're ending day one here in the press room down the lake house Day two tomorrow a lot of coverage, but we're ending the day with a great interview founder and CEO of Lamani Sharon's out CEO and co-founder great to have you on love the entrepreneur angle timings great You guys are in a great position can't wait for you to tell your story. Let's start with what you guys do I see AI is hot. We're hearing it out on stage Every comes gonna be a data AI company you get the perfect What is AI? No kidding. Yeah, very hot right now and what Lamani does is we offer Infrastructure for any company so any enterprise their engineering team can train their own large language models their own LLMs using their own data and their own secure infrastructure You know, I love about your story is is that when you have entrepreneurial energy coming together You see all the people in the in the keynote a lot of younger demographics a lot of developers You start to see that movement where there's validation for the folks in the AI field scratching that itch for for years And then this wave just clicks in it just hits perfectly, you know chat GPT opens up the eyes Oh magical blog post is written But but it wakes everyone up to the mainstream of what a lot of insiders have known for a while that there's a Magical inflection point coming around software engineering intelligence metadata usage Neural networks and so the the next gen cloud hits data scaling This is kind of like a pretty interesting point and now the whole world knows about it But it's not just chat GPT. There's so much more. What's your vision of your company? You guys have a really great solution to help developers change the game What's your reaction to that feeling and you started a company for it before chat GP he launched You're in the middle of it. What's your view? Yeah? Well, I think the future of software engineering is LLM engineering I think there's been a lot of a little bit of you know Existential crisis for software engineers on what is my future look like when these models can write code? And my take on that is you build the models because you actually sit the closest to the data You also sit closest to the product you can deploy these things Better than anyone else better than I believe even top AI PhD is a machine learning engineers like myself And I'm building the tools to enable these people to do that Remember back when I was a small young child in school I had to build the taxonomy and the seed and the the data and there was a lot of hard work I mean it was so grueling. Yes, and now it's automated. You have all these tools LLM brings a whole nother level of convenience speed agility intelligence to that Important provisioning what do you want to call a process? This is where the magic is a lot of people are coming out and saying This is a game changer What is LLM engineering because this is kind of where the sweet spot is all the heavy lifting can be just done for you now If you're a developer no more grunt work or whatever you want to call it the toil the Undifferentiated heavy lifting in this case differentiated lifting. It's a lot of pain goes away Yes, so to talk about what that pain looks like so I think there are two, you know, very popular LLMs now in production that people are familiar with one is chat GBT and The other one is GitHub co-pilot and these models, you know They were trained on top of what's known as these foundation LLMs foundation models that have been out for actually a couple years You know, I've been out for a few years so since 2020 which was a groundbreaking year for AI researchers But no one really no one else really cared but it took a team of Hundreds of top AI researchers at open AI to be able to build both of those and be able to train them Over several months and even you know years of just like waiting and being able to even gauge what What they had to build and it's making that process much simpler such that everyone can have their own large language model Their own LLM just as powerful as those models on their own data and infrastructure That's what we're building and making it possible with you know our tool But you know They're a bunch of other tools that make this possible too and it's very exciting to see this LLM engineer Emerge to be able to steer these models towards those real end-use cases towards those real Problems and in effect effectively end-users so that like have the magic touch everyone What's interesting about LLMs? I'd love to get your thoughts on this because this is something I've been thinking about large language models Is what it stands for but it's also math involved. So it's not just language. It's code to there's other data So this idea that data is the value and with open source booming. I mean software is almost I mean It's free so that so it's not so much just the software now It's the data and the software and AI kind of together This is something that's come up a lot in conversations around. Okay. Yeah, I get open source I get open AI they strip mine the web. They got it all out there Whatever great large language model, but I have my own data. I want to build my own I don't actually want to commingle My LLM, but I want to take advantage of the the big the big ones the proprietors What do they call them old word proprietary, but say open AI and chat CPT? Yeah Why not leverage it for syntax for formation for formatting, but I don't want my data to go in there So it's a lot of focus on this piece. How do you look at that and how's your company help the developer get? Through that kind of a not hole if you will it's like because it's kind of one of those things that comes up a lot Yes, and a quick note on the language thing. That's largely historical That's where you know these models first emerge from it doesn't have to be necessarily language I mean the most popular ones are all language-based because it understands English and knowledge and around the you know data component the most important thing about LLM engineering is understanding the data so those who understand the data will be able to Handle and be able to build their own LLMs the best And so those who are going to be at the forefront of the space are those with the best data It's clean. They understand what the data actually is which by the way a very tall order For some companies, but having their data act together is very important I think the second piece that is really important is Understanding the product like understanding what your end LLM use cases Chatchi BT. I know it took the world by storm last November But before Chatchi BT several months preceding it there the model is already out a very similar basically the same Very very similar model that researchers knew about but no one else knew and what differentiated it was largely just the interface It was just an interface change putting us chat interface was very accessible to people And so that's what did it and so I think those are those two things data and this like product component What do you think the discovery process going to be for the developers out there who want to have their own LLM start getting into this They're learning what to do with their data if they know their data and they're somewhat know their outcome I think I will admit that we have a lot of data We know it well, but we really don't know what the value proposition is. Yeah, I can query it But like that's hard to like so you got to play around with it So how do you see that? How do you talk to people when they say, you know, help me figure out my LLM strategy? I know I got to do it. It makes sense. It's kind of a no-brainer But the definition of the outcome might not be clear Yes, yes, I think there's a lot of excitement especially top-down and also bottoms up but I it's funny like the excitement is at the you know very top in the very like bottom of us and I Think to understand what the value is I always push companies to understand What is the business value of training our LLM? You know what what are you actually getting out of this and I think for many companies? They've already explored it's very clear it affects their top line in this way it gets more customers For example, it brings magic to their product for other companies It's largely more of a marketing play and for those I kind of want to encourage like how how do you further this into your? Into your narrative. What is the future of your entire industry if this technology is possible and I think you know one of the most? Important things when you're getting started is just to you know, think of some of the more, you know clear use cases I think one like very clear use case that we do See with a lot of customers like start with something small so start something clear Maybe it's just chat GPT on your private data. It's just question and answering on your private data. That's it That's something more, you know tangible you've already played with chat to BT It's approachable to do that or get a co-pilot a co-pilot on your own source code That's it on your own DSL your domain specific language And so just just those two applications themselves gets you started you have something working internally and then boom You can take it to the next level think of creative new ways Of using LLM. That's a great that's a great roadmap. Thanks for sharing The other thing I want to ask you I love your thoughts on is The consumption side and I see two things emerge and want to get your reaction One is companies who see the benefit AI and they bring it into an existing thing and make it better And then you got this entrepreneurial mindset of i'm going to build something native like from scratch What's your view on that? I mean it's kind of forming you seeing a lot of entrepreneurial activity opportunity recognition is clear Company taking AI in make it better and then hey, I want to go get some of that that land grab out there It's AI and grab this going on. It's legit, but it's happening. What's your thoughts? Right. I think a question that I see a lot of companies asking across You know expertise in AI is how much ownership do I want on this? You know layer for LLMs. Do I want to own the entire model? Do I want to own like all those AI engineers, you know and researchers that open AI does? Like where do I stand on the spectrum? Do I only want to just call chat to BT and call it a day? Like where do I stand on the spectrum of ownership? And I think what I encourage companies to think about is, you know, the last couple waves of You know the mobile platform and the internet revolution Those were all big ones and hopefully you did get on it to some extent where you have software engineers right now And you maybe have mobile developers if it's relevant to you But just encouraging people to think about, you know, what level ownership makes sense to you? And I think what we offer is similar to actually what data bricks and others offer as a database engine But instead of a database engine word LLM engine and we're just infrastructure And so I think what customers, you know often think about is Oh, um, how do I build my own AI mode? Do I have to also build that out or will that infrastructure piece be More of a service across a lot of different companies and that's our bet So like our bet is LLM engine is just like a database engine No company should be able to no company should have to build their own But they can still have an AI mode to using this engine and what that AI mode exactly is Is their own understanding of their vertical their own understanding of their space codifying that as an objective as What we call an objective or reward to the model and just being able to say like this model should say I don't know in these situations this model should not hallucinate about data in this way and this is situation And like those that itself is the AI mode for every single company in their own verticals because they understand their space best And all we do is run the optimization and that's their opportunity to be differentiated have that mode That's their upside. That's their that's their upside. Talk about your business. Um, for a second Give give yourself a plug here. So when I get the word out You guys have an LLM platform of every developer to build customized private models Easier faster better performed than general pros LLM you mentioned that Are you I say wait, I see a waitlist down there. I see an API Are you doing more professional services to bring the customers to prime the pump in as you got this engine? How are you guys engaging right now with customers? How do people get involved with your company? Yeah, great question So we have a self-serve tool that you can use today and that thousands of developers are already using today and we also have Basically different levels of self-service. So we have an early access enterprise program where we engage with different customers You know at an enterprise level and there are like a little bit of professional services Our goal is to make this tool entirely self-serve So the goal is to make this super easy. It's in your own infrastructure Right, like you don't want us meddling in there. So it's all yours. You can take it All we do is help manage that compute It's nice to prime the pump a little bit of get the early customers in to show the use case and then go flywheel Talk about the original story. I love the story. You you mentioned before you came on camera It started before the big hype with chat gp. So you kind of were in the field for a while Talk about the original story how you guys came together on this How did it all come together was an issue or scratching you made an ai talk about how it all came together Yeah, so most recently I was computer science faculty at stanford teaching generative ai I also teach one of the largest course error courses in the world also in generative ai continued to teach Um, also did my phd in generative ai at stanford I advised by andrew aang so i've been in the space for a very long time almost a decade And my co-founder has actually been in the space even longer than me almost two decades. So uh, we came together He had previously um launched a large language model and production at by due to over a billion users and so it's one of the largest out there and uh, you know, we came together and what we found was Wow, there is a people are really struggling at doing something that for us like We've actually both built like custom systems like this Many many times like 50 times together, you know, and It's simple for us. It's like part of my whole phd, you know, I just this is your world Yeah, I've just been doing it and There is a lot of infrastructure involved. It is not easy to get this up even for me. And so just thinking about this Um, how do we make this easier for more people? Especially people who? Actually understand those end user problems best and how do we get those people to start building these things because that would be Way more magical the llms that we would get out there right now We only have a few that are you know, I think already super magical But to take it to the next level on everyone's data and for every single possible use case that's not going to be um That's not going to be brainstormed by like only ai researchers. No way like it's not going to just be ai researchers in silicon valley And I think that's so revolutionary because one of the things that you just brought up was is that how hard it is but also the magic of I'm we're kind of a data where we know all the videos we store all that metadata But we've got a little llm opportunity there every company has a lot of data They need that expertise to end the software the mechanism to put it together To kind of take it to the next level. I see this as a revolution in software We've been calling the term data developer because we see a future where The stack will change a little bit you see data products emerge Some algorithms maybe you have some expertise and then developers just coding in line with native data But the data warehouse world is so old and antiquity They're like it's all about lake house now right siloed Well, lake house is just the beginning step to kind of transition to a world where it's just abstracted away, right? So the magic here is if it's if you have data completely abstracted away Yes, then then the developers don't have to get involved. They're coding And if they have co-pilot kind of features I just think I've never seen anything so exciting from an advancement standpoint in computer science What's your reaction? I mean you've been in this now. You're like, so you're probably so open ai coming. You've been doing it Yes, I'm friends with the founder. You're like, okay. This is going to be big Yeah, what moment did you realize this is going to be massive like you like we got to do this Yes, so for researchers. It was really 2020 when gbd3 came out I mean gbd2 was also um already kind of portending that in 2017 Um, so that's when you know, I I know this sounds funny, but back then I thought it was already hyped up This is nowhere near what I heard. Oh my god. Yeah. So now, you know a few months ago and someone told me there's a south park episode I just thought what? I had no idea because everyone around me was already speaking this language. So it was just um Uh I it's it's already been huge pre chat gbt. My go-to-market motion was educating people about llms I didn't even really use the word llms. Even generative ai was really foreign to people That was a really uncomfortable concept. And so just talking about it as ai for language Was a much more comfortable way of going about it And I actually remember some of the demos that we showed people almost Didn't believe us because they weren't able to touch it themselves or chat gbt And so I I think um, it's a completely different world we live in now and the pace is So extreme. It's it's very exciting me a lot of exploration of innovation. Sharon. Thanks for coming on the queue I know you're super busy. Thanks for sharing your expertise. Put a plug in for the company. What are you guys looking for? You got some funding. How big are you? What are you? What are your goals? You guys hiring obviously let's take it in to give a plug for the company Yes, um, we're definitely hiring engineers that I say I'm hiring engineers Um, and one thing that we do really care about the company at the company is underrepresented groups and being able to Create a home for people like that because I am one of them one of the underrepresented groups in tech and I've been in situations that I didn't super enjoy and so I want to create a home that is welcoming To people like that and who need that. Um, so yeah doing some great work solving big problems making it easier people Thanks for coming on the queue. Thank you for having me great mission Just a cue of getting all the data and we're going to share that with you soon I'll be an AI bot on camera with mid-journey avatars Sharon, thanks for coming on appreciate it. All right. That's day one wrap up. I'm john furrier host of the queue Thanks for watching