AI with Alec. Get smarter on AI. The easy way.
00:00:00 · Alec Coughlin, Tumas Rackaitis
[00:00:00] S0: Here we go. Fireside chat. We got Thomas from Rogo who is actually an employee of Lazard. He's gonna tell you more about that, but not really. Here we go. Quick intro, Thomas. Go for it. [00:00:12] S1: Hey. I'm Thomas. I'm the CTO and one of the cofounders of Rogo. Yeah. We've been doing a lot of really, really interesting things using LLMs with... You know, and helping... Using LLMs to help bankers and people in private equity, all sorts of other, you know, smaller and more interesting places inside finance, you know, some specific types of hedge funds and so on. And it's been pretty awesome. We've been able to, you know, work with people like Lazard and other really large financial institutional firms to do that. Yeah. [00:00:43] S0: Excellent. Excellent. Alright. So we're gonna kind of just take one step back because I love this stuff. So help the audience understand a knowledge graph. What is it? How does it relate to creating the first, your words collectively, right, AI analyst on Wall Street? And walk us through an example like a case study or a for example of connecting the dots between Rogo, Knowledge Graph, and Wall Street firms becoming a 100 times more productive. Go.
00:01:13 · Tumas Rackaitis
[00:01:13] S1: Look. So a lot of a lot of finance... And historically, it's always been a... Like a... Like an apprenticeship. There's a lot of institutional knowledge. Right? There's like a... There's like... They're like sith lords. Right? The master and apprentice. Right? And you always have to, you know, learn a bunch of things about specifically about your domain for any specific type of industry, for anything like that. And additionally, like, part of that also goes into the data engineering side of this. Right? So what we did is we built, you know, companies change all of the time. There's these things called security masters. So one of our data engineers was, like, x Citadel along with other people. We have lot of exquants on our team. They built, you know, a security master, which is basically keeps track of what securities are what. Make make sure that, like, you know, when you say Maple Bear, you actually mean Instacart, all of these different things. It made it much more semantic, made it much more available for large language models to do that. So atop that security master, we realized, hey. We're so good at identifying what you mean whenever you mean energy drink companies. We know... We have... We we, you know, we have such great intelligence about understanding semantic intent, how it maps to specific companies. But then we realized, wait wait a second. This is incredible. This is probably the best way to access different parts and different types of entities, different types of, you know, nodes on a knowledge graph. Right? So what we started to do is we started expanding. Right? Whenever you find a company, we'll find all of their latest funding rounds. We'll find any type of acquisitions that they've done. And then from those acquisitions, you're able to traverse the graph of different types of relationships companies do. If Wall Street is anything, it's really just a bunch of relationships between a bunch of companies and, you know, advisers like Lazard and other, you know, large financial institutions are really the people in between. So understanding all of those different relationships is so key and important. And when it comes down to making an AI analyst, that's literally it. Right? It's being able to understand those relationships.
00:03:00 · Alec Coughlin
[00:03:00] S0: I love that answer. And let's go a little bit deeper. So Wall Street's first AI analyst. I I mentioned to you before we jumped in here, right, PowerPoint creation. Like, I loved how specific you guys were because we've all been there in in different roles in in in corporate America of how much time we spend and and how difficult it can be sometimes when you're, you know, young person early in your career and trying to see the the light at the end of tunnel. And there's just so much work that feels just tedious. Right? And I'm not learning so much but I'm paying my dues and I'm doing my thing. So can you walk us through an example of of PowerPoint workflow and the way you all have thought about that in relation to what you've built and what you're building? Totally. So a way to sort
00:03:49 · Tumas Rackaitis
[00:03:49] S1: of think about, you know, PowerPoint generation is honestly an entirely different modality than what a lot of people think LLMs are. Right? You know, typically right now, LLMs generate text, sometimes images, sometimes video. And, you know, there are a lot of, you know, general PowerPoint creators. And at the end of the day, banking PowerPoints are very, very specific. There's, again, a lot of institutional knowledge there. It's another one of these things where, like, if you don't do it the exact way your partner likes to do it or the MD likes to do it, it's not gonna be okay and you're gonna be staying up till 3AM doing it and making it right. And on the other side of it, you know, PowerPoints at the end of the way are also a really good, you know, short form way to tell a story. Right? And finance is is anything is the study of what is something worth. Right? A PowerPoint is probably the most condensed information rich way to present a story of how something has value. And, ultimately, that's sort of what we've been trying to model using a lot of, you know, experts in finance. Basically, if you aren't super technical at Rogo, you've had to work in finance. You've had to work in, you know, investment banking. We have a lot of people from Blackstone and other private equity firms, or if you had to work at a hedge And so we're able to basically build an incredibly rich set of preferences of being able to gauge whether, you know, this PowerPoint is better than this other PowerPoint, making sure it fits and gets the exact type of output that really saves you, you know, midnights of work.
00:05:14 · Alec Coughlin
[00:05:14] S0: I love that. And let me play something back to you. You you tell me if I'm I'm picking up what you just put down there properly. So if I'm one of 15 managing directors and each of us have our own slightly different style and some slightly difference might be more than slightly different. Am I to understand that that's where Robo's gonna learn rather quickly all those unique characteristics and the way I like to tell stories, the type of data I prioritize, you know, that that that whole kit and caboodle is that what we're what we're getting into here is that it learns rather quickly and enables that type of almost personalization or customization? I mean, think about it this way.
00:05:56 · Tumas Rackaitis
[00:05:56] S1: And, you know, you... Whenever you try making AI systems, you kinda wanna model how people would do it normally. Right? And the normal way people make PowerPoints on Wall Street is they find precedent decks. And, you know, anything... If you build a good AI system, it's gonna learn from examples pretty quickly. Right? Yep. In fact, it's it's a pretty bad AI system, it does. And the other thing is that, you know, traditionally, people have tried making PowerPoint automation tools with UiPath or all... A bunch of rule based systems, and they never really worked. And the reason why is that this new modality of AI allows you to be a lot more flexible. When little things don't, you know, quite work, it still somewhat figures it out. You know, different types of intentions, finding the right logo happened to be, like, something really interesting we started working on. Anthropic is a lot like anthropology. And finding the right logo is honestly like a ton of work in finding a PowerPoint, especially if you have, like, a bunch of logos on a PowerPoint. That can save you hours of time, and we'll just be able to find them. We have a lot of good computer vision verification. Going back to the company knowledge graph. Right? What do we store? We store really good PowerPoint friendly logos with every single company we identify. Right? So it's this really good powerful tool that we built on top of it.