Building an AI-Native Platform to Take On the Bloomberg Terminal — with Aidan McConnell

SPEAKER_01: Most organizations are not behind on AI. They are behind on the thinking required to use it. I am Arnand Sapat Naik, ex-Microsoft.art AI executive, CEO of TipSora, and your host. This is What Comes Next. The podcast where we talk about building the kind of organization that actually wins in an intelligence-driven economy. Let's get into it. Today, for the first time, we have a guest in the show. His name is Aidan McConnell. He is founder and CEO of QuantLink AI. He has also been a data scientist at CESEL.

SPEAKER_01: His mission with QuantLink is to become the premier vertically integrated data and AI platform for the global market and investing.

SPEAKER_01: So, Aiden, welcome to the show. Thanks for having me, Run. So, Aiden, um when we met, uh I talked to you briefly that I wrote a book, and my book is based on my strategy for sustainable success. And I think there are five core stages where a company goes through to build that sustainable success. The first is the find your nest, you know, find out what you can do that others find hard to replicate. Then you start to diversify, then you become a platform so that others can build on top of what you have, and then you start building your team with ownership mindset.

SPEAKER_01: And then you start your RD, think about how you manipulate the external environment, right? Uh so that you have not only do you have long-term sustainable success, like by the time you build your platform, you have kind of built a sustainable success, but having that um RD and managing external environment is going to help you um stay relevant no matter what changes next, because you are then driving what changes, right? Now, you clearly, you know, since you started this company Quantling, you clearly have found your Ness.

SPEAKER_01: So you have achieved the first step at least. So can you walk me through your background, like you know, what you did, uh at what point you realize that there is a market for this because you know, trading and quant is kind of a saturated marketplace, right? There are big players like the Merrill Lynch, uh, you know, Charles Swabs of the world are already there, right? But you saw something that made you realize that there is a potential for improvement. There is a sector of the market that you can serve, right?

SPEAKER_01: So, you know, walk me through your introduction of you know, who are you, how did you get to where you are, and what was the idea behind Quantaling?

SPEAKER_00: Yeah, so thanks for having me on. Um I started back in 2019 or so as a data science intern at a tech startup in the Minneapolis area called CEZL. They uh do uh consumer lending. They called Buy Now Pay Later, which um is a rapidly expanding market. And when I started, everyone was in person in the Minneapolis office five days a week. Um and there's probably 50 or so people at the company, so very early stages. Uh, but then COVID hit and sort of everyone went home, and we became sort of a globally distributed uh workforce where we started hiring people in Latin America and Europe and India, all over the place and all over the US.

SPEAKER_00: And I sort of worked my way up there uh to manage a data science team. And then, you know, for the last year before I left to found my own company, I was uh the head of the AI and machine learning department, which, you know, probably had about 15 people that I hired, um, each of them, you know, from all regions of the world. So I kind of got to see uh a lot of the data science and data operations side of the business and side of finance. Um, and then, you know, during the later stages, started developing a lot of like AI and Gen AI um products on top of it.

SPEAKER_00: So it was a really, really good experience. But I was always passionate about investing and uh did a lot of trading myself and found that there were some really big gaps in the market. Um, if you want to pay top dollar to Bloomberg Terminal, you know, it's$20 something thousand dollars a year and a lot of times is overkill for the type of analytics and reporting that you need. So we wanted to build a platform that allowed people to get really premium data and be able to do a lot of the data science and AI on top of that data in a way that uh my mom or you know, a regular person could hopefully understand.

SPEAKER_00: So trying to really simplify the user experience for people is a core goal of ours and giving people a lot of leverage with uh, you know, literally terabytes and terabytes of data, but only showing the data to the agent or to the user that they really need to make informed investment decisions. So we focus a lot on the user experience and we focus a lot on the data, and we think that they're tightly uh integrated with one another. Um but yeah, that's the the general goal of the platform. Um I'd be happy to go into any more details.

SPEAKER_00: Yeah. Yeah, yeah, definitely.

SPEAKER_01: Like, you know, uh I I think what differentiates you is that like, you know, back in the day, um I I used to be on what used to be called the uh TD Ameritrade, which is now acquired by Swab. So like I used to do trading there. And what I liked was like they had a trading platform that you could place trade, but they had all the analytics, right? And they had an education course around it. So you learned how how to recognize those signals, and then you know, you you know how what options are, how do you trade this, how do you trade that, and what are the signals for all these different assets.

SPEAKER_01: So you learn and then you do the trading there, right? But like I couldn't manipulate the data. I couldn't say that I want data from this source, I want to, you know, uh kind of chunk it this way, and then I want to create my own signal and decide, right? And there are like right now in the investment market, there are like the Merrill Lynch and swaps of the world where you know you give your money, there is a money manager, and they do their own trading and they look at their own analytics and they do everything for you, and they charge uh their commission, of course.

SPEAKER_01: Uh and then you have on the other side where it's done like you know, the retail traders, right? Where you know there are these AI traders where like you know, you trust the AI, just give it to it, it does its own trading, and there are people who are just like just selling signal. Here is the buy signal, you know, and then when you get the signal, you buy or you sell, and you know, like those are kind of controversial because if you have a large number of customers and you are telling everyone to buy, you are telling everyone to sell, then that by default is kind of you know creating a new market signal, right?

SPEAKER_01: So it's not necessarily that you are getting a good uh analytics, but your platform kind of flips the game that you know you are letting people create those signals, right? You are letting people analyze the data their own way and figure out what strategy to implement, right? And on the downside, it's kind of like the people have to be really uh proficient in knowing uh kind of a little bit of quantum and kind of knowing what these financial assets are and what analytics they should be doing to get the right signal.

SPEAKER_01: Kind of like that responsibility is on them. Like, do you see that as your drawback?

SPEAKER_00: Um we started off by kind of going to consumers, and we recently, I wouldn't necessarily call it a pivot because the tools are all the same, but shifted focus more to the B2B side. And with that comes a little bit of responsibility from the end user that they do know uh, you know, they have some basic financial literacy and they do in some cases know how to manage a portfolio and to manage risk. But the goal of the platform is to, you know, basically have a lot of templated flows for people so that, you know, for the user that maybe doesn't understand all of the little nuances of doing portfolio risk management, that, you know, some of that gets abstracted away from them.

SPEAKER_00: And, you know, we can sort of help automate the process, but clearly explain, you know, what's your risk reward profile for a given investment or a given um portfolio construction? Um, but you know, our goal is to just build the platform that gives people the tools to do all of these things. So at the end of the day, the user, you know, does have some responsibility on their shoulders for making uh the investments with the tools that they're given. And, you know, there is a drawback on that that we're not just you know giving people signals and saying, you know, trust us.

SPEAKER_00: Like we're basically I would I wouldn't call it exactly open source and that you can clone it from GitHub, but we're basically exposing all of the code for any calculations, any data fetching that we're doing from any of our data vendors or data sources uh directly in the front end so that you know if you really want to go and drill into the process of how a number is being calculated, how a model is being called, you know, literally the library that's being called from GitHub, uh, you know, whether it's TensorFlow, XGBoost, no matter what, you know, we're showing people the actual code that goes into creating the algorithms or creating some statistics so that they can trust the process and even change some of the things themselves.

SPEAKER_00: If they, you know, want to get an AI to write its own algorithm or write its own functions or to even just manually edit like they did in the old days. Uh, you know, they can do all that through our platform. And so, you know, we want to, you know, give people that flexibility, but I think a lot of people probably won't use that. But it's good for building trust, I think, because we don't have uh the liberty of, let's say, building up 10 years of institutional credibility. So I thought it would be a very good idea to build trust by just exposing uh literally all the code that we're using to fetch data, to make any transformations to the data, to create models.

SPEAKER_00: Um, even our agentic systems we're going to basically expose so that people can create their own agentic systems if they you know don't want to use ours or they have their own uh investment goals, niche investment goals that maybe we didn't handle with uh our defaults. So that's kind of how I think about that. But yeah.

SPEAKER_01: Yeah. I mean, you know, like once you understand, like once you find your NIS and you understand like what is it that you want to do, what what's your uh differentiated value, I think the first step is to you you build your core, right? Like you you mentioned agentic system. So I assume like you have already kind of built your AI layer that you know that is kind of you know your operational efficiency is kind of built in. So then you would add your human layer. Okay, with that's kind of the AI philosophy that I kind of teach that you know, if you are a startup starting today, once you have your business idea, first you build the AI layer, let AI do as much as possible, and then you add the human layer on top of it to kind of you know make your decision and strategy engine, right?

SPEAKER_01: So walk me through what type of operational efficiency you have built in through AI or otherwise, that you are kind of strengthening the core of the business so that, you know, down the line you can build on top of.

SPEAKER_00: Yeah, I would say, you know, all of our software, like 100% of it, is is written by AI. But, you know, in the early days, we found a lot of gaps with uh AI is hallucinating or not following uh rules and you know, guardrails that we've set up. And so we have developed uh basically an ontology which has strict validation, so basically past failed criteria that's deterministic. So a big problem with AI is or Gen AI is that it's not it's a non-deterministic system. You know, you can uh have several users prompt the same thing and then you get different outputs.

SPEAKER_00: And so we think that gives a lot of flexibility to build whatever you want. But at each layer, when it's calling a function, we basically have an ontology that links everything together, essentially in a in a graph that uh has a strict validation at each step. And so this helps us both accelerate all of our development of the platform through the use of AI, but also make sure that uh, you know, each data point is rigidly validated in real time before anything goes to the user or anything goes to our code base.

SPEAKER_00: And so um, you know, I think that's what a lot of the bleeding edge AI companies and startups are doing these days is developing ontologies. And so we think a lot about how the front end should be represented of ontology, the data layer should be represented like that, and you know, anything in between. So that allows AIs to, you know, build whatever we need to build, but each step of the way gets strict validation. Um I don't know if that necessarily answered your questions, but that's sort of the core, I guess, technical idea behind our backend and all of the systems that we're building is to have sort of this graph architecture where everything is documented.

SPEAKER_00: If you look at the SEC, they have something called uh XBRL that's embedded within all of their filings. So if you click on one of the hyperlinked numbers, it might just say 20 million for revenue or something like that. It's hyperlinked on the SEC website. And when you click on it, it pulls up pages and pages of documentation surrounding that single number that the user sees on the front end. And so that's a, I would say, primitive and very robust version of an ontology that the SEC built. Um, but that just handles sort of the data or the metadata about the data.

SPEAKER_00: We sort of do the same thing with data, but then we do the same thing with front end. You know, how we define components is yes, the implementation might be in React or Next.js or JavaScript, but all of those components are defined in JSON and hyperlinked so that the agent, the user, all of our deterministic systems are operating off of the same system. And, you know, that really allows us to accelerate. It might sound like it's cumbersome, but AIs, if they're not constrained uh by something that's sort of like a ground truth, they will hallucinate and create slop, you know, as people call it.

SPEAKER_00: So uh yeah, we just think a lot about systems engineering and how all of these things are related and how we use AI uh in a way that's really thoughtful and doesn't really slow us down, but speeds us up, I would say, you know, probably 10x.

SPEAKER_01: Yeah, yeah. So so you you have built your uh responsible AI framework or like uh you have built your uh metrics to evaluate as the AI systems are working, right? So you know in real time, you know, how it's performing and you can apply your um improvements, right? So can you walk me through like what's your AI harness like? You know, how like down the line, if you realize that okay, this model is not performing, maybe there is another model that can do this better, but you could kind of swap the model within your framework to kind of you know improve that one part of it and you know get to the better system?

SPEAKER_00: Yeah, so I mean we basically built everything so that's interoperable uh among different models and agents. Um so that's why everyone's talking about the harness. Uh you know, we think first at the layer of the data, you know, when we're evaluating models. And so that's somewhat of a deterministic um a deterministic test or suite of tests or evals is the word that everyone uses. We have a set of evals for just data fetching. And so we have maybe, let's say, a thousand canonical data sets across our different providers, and this could be something as simple as income statement, but everyone has a different representation of income statement.

SPEAKER_00: Uh, the SEC might have one way of standardizing it. Uh, you know, our other data providers like NASDAQ, you know, they might have a different way of representing the standardized income statement. And so we have maybe a thousand or so canonical data sets. And so if you just give the agent a thousand data sets to search from, they're going to come up with uh most of the time pretty junk answers because it's too much, too many things to traverse. So if you look at like we we talk a lot about Amazon's taxonomy, Amazon has something like 20,000 canonical bottom level uh categories, product categories, but at the top level, they might only have like eight women's clothing or clothing, you know.

SPEAKER_00: So they think a lot about taxonomies and how you should structure the categories of systems or things that you can select from or traverse through. And so we define these taxonomies and basically have the agent and the user both operate off of the same system. So it's not like, hey, you search through all the bottom layer 1000 data sets, you go to fundamentals and then you go to financial statements, and then you go to income statement. And so we build the a structured set of tests or evals that you know generate realistic user questions, you know, fetch me Apple's income statement or fetch me revenue or net income.

SPEAKER_00: Things that we know or are pretty certain there is an exact deterministic answer for. So fetch income statement, fetch this column or this metric from that data set, and build a deterministic test suite just on the data fetching layer. So that to me is the most important thing is are they fetching the right data set and then are they fetching the right records or uh let's say in the financial dimension, it's period and fiscal year. Are they selecting the right thing from that data set? And we build you know thousands and thousands of deterministic tests from that.

SPEAKER_00: Then there's the operating layer on the product, which is like everything we're doing is with one of those data sets. On top of that, you can construct a portfolio, or maybe you build an Excel model in our like uh Excel like front-end feature, or maybe you're building a Jupyter notebook, or maybe you're building a screener, or maybe you're building a chart. Whatever the product layer is that the end user wants, there's another set of evals for which are, I would say, less deterministic and a little bit more of an art where we're doing sort of qualitative and not quantitative evals.

SPEAKER_00: Was the chart something that looked institutional if that's what the user wanted? Well, what does that actually mean in terms of validation? That's more of a personalized eval that you know, we sort of have to be the ones that determine whether or not the answer was good, whether the output was good. And you could do some maybe like uh visual machine learning, some CNNs or something to try to match and see like, does this Morgan Stanley chart look like the ones that our agents outputted? But at the end of the day, it's a little bit more of a personal opinion.

SPEAKER_00: Uh, did this chart look good or bad? And you know, when you have enough users eventually, that's something that you can um basically do sort of a a loop of feedback where you take, hey, did this user whatever the signal might be, maybe it's they activated after their 14-day free trial, or maybe there was more engagement after seeing that versus less engagement with whatever the the counterfactual or the B test if you're A B testing these things. So um that can be more of like a long-run goal where like, hey, did this user convert, did they churn or not?

SPEAKER_00: Uh, versus like the data fetching layer, which is where we spend most of our time, is a bit more deterministic. I mean, there's still some nuance there, but you know, when when a user asks for revenue, you know, there's should be one chain of of data fetching logic in the DAG, basically.

SPEAKER_01: Yeah, yeah. I mean, as long as you built the framework that you know you can down the line segment and uh build your A B testing to have that continuous improvement, then you are on your way to kind of you know improving your customer experience uh you know as you go, right? As you onboard more customers. Now, let's move to the next uh part of the phase. Now say that you know yeah, this is great, but you know, it's it's a good idea. But like I said in the beginning. Beginning. This is also a very saturated field.

SPEAKER_01: There are a lot of competitors, and other people can come in very quickly.

SPEAKER_01: So what I always say is like while you have the edge, while you are kind of the only one in the field and it's working for you, you need to think about diversification, right? So while you are providing data science capability to people who want it, what are some other tangential things that you could be offering that that would kind of diversify your portfolio and strengthen the core? Like I think, you know, we talked about building that font course for, you know, um, you know, maybe universities or people who want to learn themselves.

SPEAKER_01: What are some other services that are you that you're thinking of that are tangential to your core service but could become like a different revenue stream, different product on its own?

SPEAKER_00: When I initially started this company, it was the data science and the AI that I thought was going to be the biggest selling point when we were going to market. And that's actually not the case. Uh that stuff's great, and I think is going to be the bigger opportunity long term. But most of the people that we talk about, what they really love is our charts. And you know, what we essentially built is sort of like a Tableau-like platform within our BI tool, which is just a subset, but kind of feeds into everything else in the product surface.

SPEAKER_00: Where we, you know, we we determined that a lot of the existing BI tools, their general BI tools like Tableau or Power BI, are really more for internal reporting and internal analytics. And, you know, the existing platforms in our market were more of like, hey, build these templated charts with our UI and our branding. And those charts look good, but they all kind of look the same, whatever that platform might be. And we sort of operate in this middle layer where we allow you to import any of your data sets.

SPEAKER_00: So yeah, you can use ours. We have a lot of data sets, but you can import and connect uh your data warehouse, whatever vendor API keys you have, your Excel files, your Excel models. And, you know, we have sort of self-hosted storage, cloud storage that, you know, you basically operate like a uh folder and file system on your computer. And then the ability to give pixel level control over the output chart is something that there's actually a ton of demand for, as we found. And every time we're presenting, like that's what people kind of double-click on is like, hey, I can build a branded chart with our logo, uh, with our watermark and make it look as great as I want it to look, you know, something like a 10K presentation that a public company would give.

SPEAKER_00: You know, we basically allow people to build any sort of chart you would ever want to build uh on top of your data or on top of our data. And so we think that that wasn't my initial, you know, when I was building the platform, I didn't think that that was going to be the thing that would have a lot of demand, but a lot of people are really focused on the output and the presentation layer. And so building sort of a BI platform, a tableau, if you will, for this market specifically has had a lot of demand.

SPEAKER_00: The other thing is reports, which are like, you know, the PDFs, the word processing, you know, boring stuff, but that's something that people really like is having the ability to build uh their branded reports that they can share to external clients, you know, obviously backed by the data in the charts that you know I've already talked about, but you know, the end result is some external communication mechanism, uh, you typically a PDF or a PowerPoint or you know, some charts. Um, that's where there's been a lot of demand for, not just, you know, hey, analytics and machine learning and and AI.

SPEAKER_00: So we found that that's kind of like where uh a big wedge in the market is. And also, you know, our go-to-market strategy right now is going after universities because they've sort of been neglected by the big players that they use. And there are a lot of like compliance and approval workflows that are, you know, the really boring stuff of what they do day to day, but they're spending a lot of time, you know. Hey, if I wanted to buy 30 shares of Apple, um, you know, they're documenting this in emails.

SPEAKER_00: And, you know, the approval chain is like this interpersonal frictional process where, you know, it's a pretty simple thing for us to add, you know, it might take a couple of weeks, but it really removes a lot of friction, at least on the university side. And I assume it'll probably expand when we're um, you know, going to money managers who have to deal with the same approval workflows. Another big thing is they found that they were messaging over WhatsApp on their phones or emails, which they didn't really have a hub that was specific to their fund group uh for universities.

SPEAKER_00: And so we built kind of like a Slack clone within our platform that allows people to just easily message people, create different channels. Maybe a lot of these funds groups and universities are kind of structured where they have teams that are oriented to a specific sector. So they're focused on the tech sector, or maybe they're focused on financials or something like that. And so like having chant the ability to create channels, Slack-like channels within the platform so they can just kind of have all their workspace stuff, all of their messages, all their approvals, all their data kind of in one shared workspace.

SPEAKER_00: Um, those are things that I kind of thought about early on, but just from the conversations we've had, there have been a lot of demand for those sorts of features that are just the general, you know, Google Drive type of features of shared files and messaging and all that kind of stuff. So I think that's where a big opportunity is, is just doing the boring stuff really well. Uh it's not like the really exciting stuff that's gonna get a lot of clicks and a lot of like focus, but um, it's something that there's real demand for from the conversations we've had.

SPEAKER_01: Yeah, yeah. That's good. I think, you know, you you mentioned like you know, you're taking the Google approach, like start with this, but then you know, also have the drive and the the the the custom charts and all. That that's great that you're already thinking about diversification. I think you already started with the concept of a platform right from the beginning, which is great because usually like you know, that's kind of uh the third stage that we think about. Like, you know, you already have a platform where people can, you know, analyze data.

SPEAKER_01: Like, do you think about kind of generalizing the platform enough that people can down the line decide like you know, there can be like marketplace of data providers, there is marketplace for like models and you know analytics or those signals, right? And you know, marketplace for reporting, marketplace for storage, right? And marketplace for like the Slack channel and the messaging that you are talking about, right? Uh uh on top of that marketplace for agents that can act on these signals, right?

SPEAKER_01: Like, are you thinking about building a platform where it can be a plug and play that other people can you know bring their own agent to it? People can bring their own data to it, people can bring their own, you know, quant analytics of models to it, people can bring their own, you know, uh reporting and uh charting capability to it. Like I uh are you thinking that in your long-term vision?

SPEAKER_00: I I wouldn't say long term. That's that's very short term. And it's funny that you brought that up because uh that's what we've been talking about every single day for the last month, basically, is shifting from uh you know kind of a closed platform where it still is a platform, but we're the ones that determine the data and we're the ones that determine the charts to something that is just more of an open framework and a marketplace where people can you know bring their own API keys, they can sell their own algorithms, they can build their own agentic harnesses, um, they can build their own chart templates.

SPEAKER_00: You know, if you look at like Figma and some of these other platforms, like a lot of it, they are kind of marketplaces where people can uh you know build their own templates if it's for design, or maybe it's uh Snowflake is a really good example of a marketplace where data vendors will just kind of plug in there and you know they can sell their data through the the through the Snowflake marketplace because they have so much uh scale and distribution. So, you know, building uh the set of tools that allow people to build any chart or write any arbitrary code, and we just are maybe the storage layer and we give people the tools to build sort of whatever they want, whether it's front end or back end or data, is really what we you know have talked about literally every day for the last month.

SPEAKER_00: So it is funny that you brought that up. Yeah, that's that's what we're doing. And I I wouldn't say long term, I would say it's you know, within the next couple of months we'll we'll have um I guess a great, great, great.

SPEAKER_01: Yeah. So uh you know, l let's talk about like the people structure. Like how many people do you have in your company right now?

SPEAKER_00: I think including me, there's ten now, um, all working in our Minneapolis office. Um and you know, pretty heavy on the technical side. We have two product designers who are great, and then you know, the rest of the people are mostly technical, and then we have uh one person who's kind of leading sales, um sales and marketing. Uh, but yeah, the rest of people are AI uh engineers or quants or just developers. I shouldn't say just developers, like uh they're they're all they're all really great. And you know, I've I've had uh kind of I would say recruiting is the hardest thing, at least in the early stages, because I've had to you know convince people who could be working for you know Google or Meta, you know, to move across the country to freezing cold Minnesota in the winters.

SPEAKER_00: Uh huh. Summers are nice here, but the winters are brutal. So, you know, convincing people over LinkedIn messages to, you know, give it a shot is uh you know where I would say you know, the bulk of my work was early on and and um you know really convincing people on the vision of the platform and you know what our mission is, um I think attracted a lot of people to apply for the roles and you know get some really talented people to come in the door.

SPEAKER_01: Yeah, yeah, that's great. So I and how have you created your incentive, we talk a lot about building uh your people structure with an ownership mindset, right? So do you provide equity or like how much flexibility people have on like, you know, what how how they shape the product or you know, do they get time to kind of cre build their own product and make their own case for, you know, uh including it within the product roadmap or creating a side product, like how much flexibility do people have in that way and what level of ownership do people have?

SPEAKER_00: Um on the equity side, we we give people RSUs. So I'm a big believer in, you know, like if if the platform or the company wins, like they should become, you know, very wealthy as well. Um I I was very blessed to get in the door at a at an early stage startup and watch it scale to a multi-billion dollar platform. Um, so I know, you know, kind of the risk reward profile of working for startups. And so, you know, giving people equity, I think is very important to incentivize them. And then, you know, on like sales and marketing side, really rewarding people if they're bringing business in the door.

SPEAKER_00: Um, and then on the product roadmap side, you know, I think everyone contributes to the roadmap. And I allow a lot of flexibility because I'm not gonna have all the answers. And so people will come and, you know, you can vibe code a demo pretty easily now. And so like ideas that you have you can implement and see a front end attached to it uh, you know, within a day now. So like we we allow a lot of that. And sometimes it'll be something that I didn't really think of or no one else really thought of, and we'll have an intern who you know vibe codes something that's you know becomes a primary product surface within the next couple of weeks or the next month.

SPEAKER_00: So yeah, we we think that's very big is you know, we we want to stay very focused on the things that we're developing, but you know, giving people that 20% time to to really experiment and learn how to use AI, but then you know, how does it apply to our market, to our customers, to the problems that they're having? Um, that happens all the time. You know, every month it happens.

SPEAKER_01: Yeah, and and outside of like your employees, are you thinking of building like an affiliate network, commission-based sales, advisor network who are not your employee, but you know, if they bring in business, they have some incentive?

SPEAKER_00: Yeah, we we haven't set up the structure for this yet, but we have talked about it a lot. Um yeah, we we think that affiliates and commissions would would be a great idea for people for anyone who brings in business in the door. And so um I think you know, the simplest way to, I guess, get a V1 out there is, you know, if you get someone to sign up and be subscribed, that you know, you get some sort of incentive. Maybe it's AI tokens, you know, you get a thousand dollars of AI tokens, or you get uh even uh a couple of months free off your plan, or maybe it's you actually get real commission like back in your bank account.

SPEAKER_00: I don't know what the the legal and compliance rules are around that, but um any way we can reward people for for building our business, you know, we should definitely do that and incentivize it heavily.

SPEAKER_01: Yeah, yeah. I mean, you know, be beyond like you know, building that people culture with ownership mindset is one of the key to future growth. Uh beyond that, you know, um I say that you know once you have that level of security, you should think about investing in research and development so that you build the next red thing that may replace you know your thing or everything else in the market, right? So if you had unlimited funding and you wanted to invest into research, what are some opportunities that you see that you know uh you think that there is uh you know um more research required here, and you think that something could be invented here that could completely change the way we think about investment and kind of build, you know, um uh financial security for all.

SPEAKER_01: Like w what what are some wish lists of where do you want to invest in research?

SPEAKER_00: I think what you're finding with a lot of these vertically integrated platforms that are operating in a given space. Uh Harvey is a good example of this, where they're you know building an AI platform for uh for the legal space for law, and you know, specifically big law, like out on the East Coast, um, these massive law firms. And, you know, they obviously have a model suite of the big AI labs, but they've started developing their own models there. And these are like custom tuned models, trained models on legal workflows specifically.

SPEAKER_00: And so I think there's a ton of space to operate when it comes to fine-tuning and training uh models within a given vertical slice of the market. And so for us, maybe it's how to produce alpha with uh transformer architectures, and that would be layer one. Layer two, I think, is a fundamental rewiring of the architecture of AI models. And um, I you know, I could talk a lot about what I think where I think this is all gonna go, but I do think that LLMs or AI architecture in the future, let's say five years, is probably not going to be outputting code.

SPEAKER_00: It's going to be outputting, you know, more of like a structured DAG graph of things that call deterministic code. So if you think about uh something simple like training uh an XG boost model, there's a structured DAG of steps that happens in order to do that. And in my mind, in the future, the LLM will be trained on what are those steps and what are the inputs of those steps, like the parameters that go into each chain of functions, and then evaluating it on the end goal, which might be hey, did this investment uh return above the risk-free rate, or maybe it's something else like you know, volatility or drawdowns or whatever you're trying to optimize for?

SPEAKER_00: I think that the code is already written for the most part, you know, all of these functions already exist, but that chain of steps is the thing that really uh is more of a boutique process that developers or people who are bobcoding with AI go and they implement, but it's really a deterministic graph of steps. And then there's an output of did this thing uh perform, you know, in in our spaces it return uh above the risk-free rate, and did it have lower volatility than the SP 500? I mean, that's the general goal, I think.

SPEAKER_00: Um, you know, that's a validation that you can run uh and you can train uh fine-tuned models that will be, you know, 1% of the size and have 1% of the latency and 1% of the cost that these you know big behemoth 1 trillion parameter models would have if they're outputting boutique code every time. So that's where I think uh if I had to guess where the architecture of all of AI is going to go, it's going to go, I mean, at least in the the digital world, the physical world's another animal. Um again the computer vision and robotics and all that stuff.

SPEAKER_00: Um, but I think in the digital world, when it comes to software, it's going to be deterministic graphs that are, you know, each step of the way invalidated or validated.

SPEAKER_01: Okay. Yeah, I mean, the the research is, you know, uh one side. Um the other part of it is the external environment, right? Like you are in a regulated industry, so there could be like government regulations that say that would affect you, right? So are you thinking about proactively investing into uh like lobbying group to have a say in um how these regulations get made? On the other side, that is also like consumer education, right? You know, people think, okay, maybe I'll just start a money market that's easy, somebody else manages, but uh to kind of educate people that they can really manage their own money and they could really, you know, like quant is a big word.

SPEAKER_01: People think, oh, I need to be super geek to do quant, right? Um so you know, uh are you investing into these consumer education that, hey, you know, AI is here now, you can now do things you could never do before. So start thinking about managing your own money. It's not really as risky as you think. And those are the two things that I think could potentially affect you down the line. Are you thinking about kind of proactive investing in these areas?

SPEAKER_00: On the regulation lobbying side, not right now, but you know, if we raise capital within the next couple of months, then you know it's something we could definitely think about. I I do I'm a huge fan of open source, and I think everyone on the team is as well. But you know, there's a lot of talks just in you know, Washington and and and local governments as well, um, in terms of what's gonna happen with data centers specifically and what's gonna happen with the non-US based models and what's gonna happen with open source.

SPEAKER_00: So um I tend to believe that you know there will be some regulatory hurdles um in the future. I just don't know which one's gonna kind of bite first. Um so yeah, we're just focused on the product and on the data right now, and and if we raise capital, that's something we can maybe think about. But um on the consumer education side, I think that's a really, really good idea and something that we just have to kind of create or at least set time aside in our roadmap to to stop for a second and say, like, hey, let's actually explain all of these concepts to users um so that they can kind of whatever part of their investing journey or or knowledge um they have, like, you know, giving them the tools so that they can educate themselves even from the very beginning of like, you know, what's an ETF and w what's the S P 500, all the way to, you know, how to do portfolio analytics and risk modeling and things like that.

SPEAKER_00: So yeah, I think uh, you know, guided workflows are kind of the answer to that, like sort of guided tutorials. And and blogs as well, and maybe something that's even sort of an interactive educational experience where maybe AI is kind of plugged into that to help answer questions along the way instead of just, you know, I don't like reading blogs all day. And sometimes I just want a precise answer to a question. So hopefully we can, you know, build a lot of that into the product itself. And then also maybe on um sort of our marketing SEO type pages that just has you know blogs and maybe some YouTube tutorials and things like that.

SPEAKER_00: So yeah, I do think that's really, really important is just you know kind of having a place where we can answer any question that someone might have.

SPEAKER_01: Okay. And finally, um, you know, uh what are some big new updates coming to your app and where can people can kind of sign up for a trial or something to try this out?

SPEAKER_00: Yeah, so uh quantlike data.ai uh pricing, there's uh 14-day free trial for everyone who wants to try it out. Um in terms of product updates, uh we're launching a lot within the next month um just to get everything out there because, like I said, we're uh doing going hard at like universities and pilots with universities, and the semester starts soon. So um our current app and what we have in about a week or two is going to look very different. But just to name a few of the things, we're adding in um, you know, a Google Sheets like feature, but with all of our data sets kind of integrated into that.

SPEAKER_00: Um Jupyter hosted Jupyter notebooks, which you know kind of allow the user or the agent to, you know, write Python on top of these data sets, create any sort of workflows that they want, um, screener with back testing so that you can back test any sort of strategy you want. Um sort of a unified folder and file structure, like I kind of talked about Google Drive, so that people can share documents, they can share charts, they can share models, all these different things. Um and then the I would say the data layer is something that we're gonna be shipping, you know, within the next week or so, which is the ability for you to import any data set, connect your data warehouse, your databases, any of your API keys you might have from a from a data vendor, and allowing you to sort of use your own data, but within our platform, um, kind of in a self-hosted way.

SPEAKER_00: Um and then yeah, like AI agents, obviously um, we're making a lot of improvements and launching a lot of things there, but these AI agents sort of just operate on top of all of those things I just listed. So um it's it's a unified um product and uh experience for the users.

SPEAKER_01: Okay, thank you. Um if this episode shifted how you think about AI in your organization, send it to one person who needs to hear it. And if you are ready to act, visit tipsora.com to get certified or reach me directly at arvnansipatnai.com. The intelligence economy is already here. The question is whether you are building for it. So that's a wrap on today's episode of What Comes Next. If this conversation gave you a new way to think about AI strategy, share it with someone who needs it. You can find everything I'm building at tipsora.com, including AI certifications for you and your team.

SPEAKER_01: Connect with me on LinkedIn by looking up my name, Arnan Chapatnaik. Until next time, build the architecture and the advantage follows. And thank you, Adrian. Um, you know, all of you know your name, people can look you up on LinkedIn and um the fontlink.ai and everything you mentioned will be in the uh description so people can uh click there and you know sign up for free trial. So thanks a lot for uh coming in. Uh, it's been great speaking to you. Hope to talk to you again soon. Yeah, thanks so much for having me.

SPEAKER_01: It was great. Thanks. Bye. Bye.

The Story