0:00

Hey, my name is Eva, and this is Blondon Quantum, the podcast that breaks down quantum technology into real-world business impact. Here we make the complex simple. No equation, no overthinking, just insight, innovation, and a bit of cure. So don't worry, you don't need a PhD in physics to follow. In each episode, I talk to funders, scientists, and investors about how quantum is reshaping the industry's today, not in some distant future. Oh and if you heard a black cat boring in the background, that's my co-host, a very alive shorting of a cat joining the conversation. Her name is Moon. Let's get started. Hello everyone, welcome again into the next episode of Blonden Quantum Startup Series. Today with me is Yani Gamvros, a CEO and the founder of Quantum Signal. Hello, Yani. Good morning. Hi Eva. Thanks. Very nice to be here. Thank you for being here with us today. Let's start then. Please tell us a little bit more what quantum signals and what problem are you solving? Right. So we are solving Quantum Signals is solving a problem that's central to financial institutions, especially to hedge funds, CTAs, also sell side institutions that are trading, that are essentially building intraday strategies to trade in financial markets. So what we are noticing right now is the use of AI and agents specifically for trading. So these are might in some cases be LLMs that are trading for these financial institutions, or might be specialized agents. And what we're noticing right now is a big split that is happening in the market. So the top-tier hedge funds are effectively deploying these very advanced models and these very advanced agents to do intraday trading. And everybody else is being left behind. So all the other smaller hedge funds, everyone that is essentially not in that top tier, doesn't have the resources, doesn't have the skills and the commitment to build essentially these uh advanced AI models and advanced agents to do this trading. And they're still potentially using technology that's maybe 10 or 15 years old. We saw recently, about um a few weeks ago in early May, late April, there was this big news announcement that Jane Street, one of these top uh hedge funds, is essentially purchasing six billion dollars worth of compute from CoreWave. And so this points essentially to this trend. It's basically already something that uh they've probably been working on for several years. It's not uh something that uh has just come up. You don't do a six billion dollar investment like that overnight out of the blue. And so it's a trend that is happening, and uh we already see that the smaller hedge funds, the smaller players are basically not able to react, right? So don't have the capacity to buy compute, uh sometimes they even lack the the detailed uh data feeds that are required from the exchanges and also don't have the skills, right? They don't have the skills, and now at this stage they don't really have the time to react to this and and try to get the skills and learn and stuff because it's already too late. They they didn't need to need a solution, they need like a package solution. Understand. Okay, and uh probably when we talk about the volume, oh, use it everyone who is not in first top 10, which means all the smaller one, which is like what 95% of the market? Exactly. We're talking about thousands and thousands of firms, and also a big part of the AUM, the assets under management that is being traded, does belong to this essentially anyone but the top 10. We're talking about potentially five, six, seven trillion uh dollars of AUM, whereas the top 10 potentially cover about seven, uh, to eight hundred billion of AUM. So it's a big part of the market, whether you look at it from a number of firms perspective and also AUM perspective. Congratulations, since like trying to capture a really, really exciting and big market. Okay, but from what you're saying, I would like to go a little bit deeper into it because you said there is a trend, everyone's going into agent and so on. Does it mean that with a time the trader job would basically disappear or work? It would be like huge hedge hedge fund, I don't know, one portfolio manager, and the rest will be basically digitalized by agent? Is that where we go? Very good point. So, yeah, so it's basically very parallel to what is happening with uh the other jobs that are being impacted by AI. Um, I think it's another tool that will be essentially in the back pockets of uh the quantitative traders, the quantitative professionals, uh traders, uh, portfolio managers, uh, the analysts, the researchers. It's basically gonna help them to explore new strategies, to implement uh new strategies much faster. The other thing to point out is that pretty much already close to 90% of all trading is done by algorithms anyway in financial industry. Okay. So we have a very small percentage of essentially what we call discretionary trading, people essentially just looking at screens and making calls. A big part of the, especially the intraday trading, but even macro training, is essentially done by algorithms anyway. Right. So this is just a next level, that's how. Exactly. It's an addition to that. Yeah, it's an addition. The one thing that we are bringing uh to the table, and the one thing that is a big differentiation in how we're doing things, is that we are essentially building the equivalent of a foundational model, right? So we are building just like you have LLMs that are trained on text and essentially are producing text in various forms. They're producing free form text, they're producing a code. We are training our models on the same scaffolding, this transformer architecture scaffolding that the LLMs are using, but we are using time series, numerical data that comes from the exchange. So this is the market data. That's what we're training the foundational model on. And then the model will be producing essentially these uh intraday predictions, and it will power these agents that are going to be used for trading. So we're not talking here about essentially producing a wrapper for LLMs that uh essentially analyzes news, analyzes text, analyzes sentiment. All these things are there and they're they're good, they're needed. Uh, but what we're doing is essentially complementary and it's one level deeper. It's industry specific, essentially finance uh native, as we call it, uh, foundational model on these time series numerical data that comes from the market itself. Okay, so that's your core insight behind. I like it. Okay, and then let's talk, let's let's step a little bit back and talk a little bit about um how did you start the company? Why that specific industry, why fintech, where did you find your CTO? How this all started? Yeah, great, great. So interesting origin story um is that both uh myself and my co-founder and CTO, Jordanis Kerenidis. We both used to work about seven years ago. We both uh got hired by one of the very first uh quantum computing software startups, and we started working there on essentially quantum computing solutions, doing research on quantum computing approaches and how we were basically tasked in figuring out what are the right applications, what are the right use cases for quantum computing in many different industries. At the time, there was no essentially AI crazy, AI adoption happening, not uh definitely not at the stage that we're seeing over the last uh essentially two, three years. So a lot of the people were actually really worried about how quantum computing will impact their industries. And we saw early on that finance was one essentially one of the key industries where people were really concerned as to how will quantum computing impact. And really, what year was that? What year are we talking about? Well, we started in uh essentially 20 uh 18. Exactly. Yes. It's pretty early. Yeah, I think the first announcement from Google on their lab uh was around well, the supremacy sort of paper by Google, the first one was in 2019. So it was basically very early on. I think IBM, Google had announced their initiatives, Regetti was uh another hardware company, and D-Wave. I think those were the four hardware companies at the point at that time. And in terms of uh software companies, basically you just had uh um QCWare, CQC, I think maybe Zapata was uh being formed uh sometime in yeah, 2000 late 2018. Around that time, right? Okay, so you were in one of the faces, that's perfect. Okay, so you saw this fear, you saw this coming from the financial institution, and then what happened next? So, yeah, so we were being asked to come in uh to these very large financial institutions, and we can name some of these names, uh, these stories are public. Uh so we went into Goldman Sachs, uh, we went into uh JP Morgan Chase, Itau, Unibanco, several others, and uh we started essentially looking at all the different use cases and all all the different hard compute problems that were being top of mind basically for these managing directors at uh financial service institutions. And basically, we're also looking at the current technical landscape within these large financial institutions. And we worked on that for uh close to five years, and we got a good understanding of where are the gaps, where are the computational gaps and where is the opportunity for improvement, and where also are the technology gaps. And so about two years ago, my co-founder and I decided to basically go about essentially solving one of these problems on our own. And that's essentially how we got involved into finance because we had worked on for many years in these uh financial institutions. We had a very good understanding of uh essentially the gaps, and we decided that we can address one of these uh gaps by uh, and it's we're better served, and the market is better served by us essentially going out and kind of building our own uh venture. Okay, tricky question here. Why would you decide? So I understand you were working on a bigger company, you had the understanding of the market. Why wouldn't you try to build it inside the enterprise that you already were? Why did you decide, okay, that needs to be a separate solution? Good point, good point. So we had a fundamentally different thesis after essentially working on this for five years. So the old way of doing things, and this was not specific to uh our company. I think essentially it's the way that a lot of the quantum computing startups, uh, essentially on the software side and the application side, um this is the approach that a lot of them were using at the time, which was essentially to look at multiple industries, multiple use cases, uh, potentially pick one or two for depending on the client, and then take the real-size problem and pare it down to something smaller, much smaller, that fits into a quantum computer, essentially build uh proof of concept, a quantum computing application, run some experiments or some simulations on maybe classical hardware or even on quantum hardware, and then extrapolate from there. So this was the typical mode of operations for, as I said, many of the uh quantum computing startups that were building applications and software. And so we decided to fundamentally change that. And that's why basically we decided to kind of launch this uh this new enterprise. So, first of all, the key thing we decided was to uh concentrate on one industry and one use case, right? Uh, I think it was essentially too distracting to look at different industries, different use cases. So we decided basically to look at a very specific problem, this intraday trading problem, systematic trading in the finance world. So we picked up the use case, we picked the industry. Uh, the other thing was to essentially concentrate on on software and concentrate on building a product, right? As opposed to, again, I think the model was back then to potentially offer services because there was a lot of exploration that was being happening on what is the right use case, let's do a whiteboarding exercise. There was a lot of back and forth there on what is the right use case and where where do we go and which use cases do we approach. But we decided to concentrate on on um on software. And then the third and final thing was to essentially start using the best tools available today and use AI to approach the real uh problem, the full-size problem, have a solution for that, take it to the market, essentially prove out that the market will react to the solution, and then slowly add quantum computing once quantum computing can essentially improve that solution. And I'm really glad that you mentioned that because I wanted to follow up a little bit more. We talk a little bit about what quantum signals are is doing, we talk a little bit how it's relying on AI, and so far we did not talk anything about quantum inside your solution, and we are in London Quantum. So I would like to ask you, I need to explain us as well. So, what is the part of your solution right now, which is actually quantum, or maybe quantum versus classical or quantum inspired? How would you define it and what's going on inside? Sure. Yeah, so as I was just saying, we just concentrate, we are initially starting by building an entirely classical AI pipeline. So we're at this stage now where we have essentially the first version of our product out. We launched in January this year. We have our first customers coming in. And this pipeline right now, it's entirely classical. It solves the real problem, and essentially active hedge funds that are actively trading, intra-day systematic strategies are coming to us and basically asking to use the solution. The next stage that uh we are already developing inside are what we call quantum inspired agents, right? So these are quantum inspired agents, so still running on GPU, still running on classical infrastructure. And they use this uh foundational model that we said earlier that is being trained on market data, time series numerical data that's coming in from the exchanges. And basically, those agents are are figuring out what is the best, essentially the best possible strategy to implement in the market and explore how to take advantage basically of the different uh of the volatility in the market. And so this is the second phase essentially that is in the works right now. And the third phase, which we have already started working in the sense that we're already doing the research. We've published uh two papers, and one is on the way on how to scale uh training of neural networks, of deep neural networks on quantum computers and how to um uh build these agents on quantum computers. We've uh we've published the research and we expect to start essentially doing some hardware implementations in a couple of years. We've already done some initial for the papers, but we expect to do essentially a larger implementation as uh essentially hardware matures. We get more qubits, more uh error-corrected qubits that we can use for this. We expect that's gonna start essentially in maybe the two, three year time frame. I like it. So you basically what you did, you you turned the table upside down. Because usually with quantum startups, which I'm emitting right now, what they usually do, they start from the research, the academia, they focus on quantum computing, and then they're using slowly as they go, um, they're looking for specific use cases and so on. And you obviously started from the end of it, from uh getting the use cases and the solution on like classical computing right now, provide it to the customers, make money, build the first traction, and then add quantum um computing on top of it. That's that's beautiful, and that solved a lot of the startups problem, how to build a traction and attract more venture capital. And I know actually we were gonna talk it about the second, because I know that you are raising right now your second round, right? Which is uh seed round 4 million euro. Is that correct? That is correct. So we've already raised uh essentially 1 million uh euro pre-seed round that was led by Quantonation. And we have we are now in the process of raising our 4 million uh seed round. Uh, we already have some interested parties, uh, some commitments through saves, but we're essentially right in the thick of it right now. Fantastic. So, dear investor, if you are watching us right now, that's a good moment to approach uh Yoni about it. But Yoni, speaking of the fundraising, I need to ask you, and then forgive me if this is a tricky question, but how do you categorize your startup because of your inspiring quantum technology? Do you really consider yourself being a truly quantum startup, or how do you present the startup to investor itself? Yeah, so actually it depends on the thesis for the investor. Sometimes we don't really go deep on the quantum piece. Maybe they're an AI investor, right? So the way that I look at it is that we are essentially quantum proof, right? So we are going, we can scale the company using AI. And then if uh quantum brings on a big disruption, which we expect that's gonna happen, our business plan, our business, our technical team, we have essentially all the tools inside to deal with that. So we're basically quantum ready or quantum proof, if you will, and we can deal with that. Now, some investors, some quantum investors want to see a more detailed roadmap and more details on our research on and what we do on the quantum side, but it really depends. I mean, the sort of the investors self-select, right? Some of them go start diving deep into the quantum side. If it's uh an investor with an AI thesis or a fintech thesis, they basically don't really uh ask those uh uh those questions on the quantum side. They want to see basically how the business scales, how does the AI work right now? That's what I'm asking because as an ex-finder, I know how tricky it can be when your narration and the story have to be different basically for a different type of investors, right? So I mean you're open to speaking with um obviously with quantum investors because you have a quantum nation on the board, so you already proved that you can do that. But you are also, I guess, speaking with the AI investor, and you're also open to general investors, DC, like I don't know, B2B, enterprise, SaaS or so on. I I don't know if it's SaaS, I just assumed. It is SaaS, yeah, yeah. It is SaaS, it's B2B, yeah. So typically, yeah, I think the the investors that like us are the are this kind, right? The ones that have in their thesis, B2B, SaaS, AI, all of the above. Yeah. Okay. And where do you where is your solution sitting in the stack? Obviously, it's not the hardware, but is it a software? Is it algorithm? Is it application? How do you how do you reach quantum signals? Yeah, so it's basically uh uh it's a SaaS application. We run everything on our side. So we run training and inference on our side, we consume the data feed from the market in real time on our side. Well, first of all, we take the all the historical data in, we train the model as we as I mentioned earlier, and then during market hours, we consume the real-time feed. We run inference, so we generate the predictions that the the foundational model generates on our side, and then we feed that back to our customers through uh through an API, basically, through an API. Okay. So does that make a lot of sense? But are you taking also data from the potential clients? Are you allowing them, let's say, because you said, okay, we're taking the the market data, which are the public exchange data, but do you have any option for the client itself to use their own data if they want it? Very good, yeah, very good question. Right now we don't. Right now, the all the data comes from the market. Essentially, uh, we have on our roadmap the capability for the client to provide some additional liquidity that they see in the market. Maybe they have access to some dark pool or some additional essentially liquidity that is not essentially public, and they would be able to essentially also uh send that over to the same model, and the model will be able to take that into consideration before making a recommendation or before running an agent, basically. Fantastic. Okay, that sounds good. And then Yenny, correct me if I'm wrong, but when I'm trying to understand your business model and your operation and so on, I see here a really good risk management from the investor point of view, because really you are in this comfortable position that if I don't think it's gonna happen, but you know, if quantum computing will never materialize, then basically this is not breaking your business, isn't it? You will still computer as as you are, and you will be completely, which of the quantum startup cannot say that. Yeah, yeah, exactly, exactly. We've built a business in a way that it's uh yeah, I guess it's quantum proof in both ways. You could mention it earlier, quantum proof when quantum materializes and becomes a big part of the the the next disruption, but it's also quantum proof in the sense that if quantum doesn't materialize, we still have an active business. But the way that I think about it is that I think we are essentially putting together a template of what a quantum business and quantum application that provides advantage, right? Or supremacy, if you want to call it that, would look like. Because I see too many of the quantum computing startups that I see are essentially producing papers on, let's say, supremacy or or advantage. And obviously, they do a lot of great research on the quantum side. What happens on the classical side is a little bit of an afterthought. And we keep forgetting that in order for quantum to really take over, it needs to essentially beat the current benchmark, right? And the benchmark is not only technical, right? But the benchmark is also on the business side, right? So I think we need to essentially prove these two things in parallel. We need to show that what is the current classical benchmark, that's what we're building right now, have that not only technically proven, but also business proven. Right, to have customers on to show that people are willing to uh spend money on that solution and then have essentially quantum come in and be able to over and do better and and have customers basically start preferring the quantum solution that does something better. I agree. I think um people who are really into quantum and a little bit on this bus and hype right now, they're forgetting that classical computer will not disappear even when quantum is is ready, right? It will still be there. Quantum will have a completely different purpose than the classical computing. That's a better, I think, uh way to say it. Uh but okay, you mentioned that yes, one of the parts of why this is important, we need to prove that people are able to spend money on it. So let's talk a little bit about it. And you told me before, I think, of the record, or maybe you also even mentioned that today that you already have your first customers. So I would love to uh to you a little bit elaborate about it and tell us who are your customers, uh, what are you doing right now, how many of them you have. Yes, yes. So we are very excited. We just launched the service in January. We already have uh four signed contracts and quite a few more that are in active negotiation. So out of the four that are signed right now and are using the service, we have one major, the European Bank. So that's on the sell side. Uh that's important because the sell side, the banks also need to essentially be running these models and these agents, right, to understand basically what's happening in the market. And they cannot always essentially build uh those on on their own, even though you might think of banks as essentially having more resources and more skills and more people, but uh, there's also a lot of competing priorities, right, within banks and the managing directors and and within those institutions as well. So that's one of our key customers. Then we have three others that are hedge funds. So on the buy side, essentially implementing these intraday systematic strategies. Actually, one of the hedge funds is uh what we would call essentially a top five quant hedge fund. Oh, yeah. They're also based in Europe. They're they're pretty major. The other two are smaller funds, right? So they might have smaller assets under management, but still kind of jumped on the solution right away. They essentially see the need to uh start using something like this right away. And so these this has this is our initial essentially cluster of four customers that are on the service right now. Okay, so this is interesting. I have two follow-on questions on this because I'm obviously not surprised that some bank decided to use your services because we all know that uh innovative teams in major institutions are usually not that innovative as we would expect. So that's a bigger pro problem. But you told me, well, you you mentioned it in the beginning that all of the top 10 most uh hedge funds they are developing their own solutions. So here you're saying, hey, but one of these top five in Europe is actually our client. So how did you convince them to use your product instead of developing their own? Yeah, I think uh I think this top 10, the top tier uh might actually uh some of them might actually buy our solution for benchmarking purposes or just to be aware of what's happening in the market. So typically these uh these top-tier firms always try out uh every new thing that comes into the market. Um and I think they're gonna keep our solution in place uh for benchmarking purposes, and who knows, maybe they will run some models on the side on our solution as well, and have them kind of running in parallel with uh their solution. So there's always this concept in um in the hedge fund world of buying essentially signals, right? So there is a big market of buying different signals from different places, combining them together, and through this combination, essentially informing your decision making and informing your algorithms and building new strategies. So we basically become one more uh signal that they buy anyway. I I didn't throw about it, like who if that not them, right? In terms of like be on top of the things and the signals all for the all of these people are signing like yearly contract contracts, or what is the standard uh terms that you have for them usually? Yeah, so right now the standard terms are just monthly. Uh month-to-month subscription. I'm happy to uh just provide month-to-month subscription. I think people will get hooked to the service and it's just an easy way to get started. And then once they're kind of really hooked, yeah, at some point I'm sure they're gonna come asking for annual uh discounts. That tells me that you must be very confident of your solution if you provide are those customers buying the solution because of quantum or despite of quantum. Yeah, I think only maybe one out of ten customers is really asking us what is happening on the quantum side. Most customers basically don't care. They just want to see a solution that works, they just want to see predictions that work, models that work. They don't really care. They don't go into the details of how it's powered, what is the technology behind. Sometimes they're just gonna ask to do some kind of due diligence on their side to see if it's something they've tried before or or you know if there really is something in the back. But yeah, that's part of kind of their due diligence. But as long as kind of it works, they don't care. That's great. Okay. And then I know this is only a couple of months because you said you launch it in January, right? But can you share any concrete example of ROI that you generated for clients, even without saying the name? Yeah, so typically the customers uh in this space are very secretive about exactly how much they use the product and you know how much they invest essentially using our product. So I don't have any information on that just yet. I I think the best we would be able to do at some point in the future to be able to say how much essentially how much AUM in general is being invested through some of these uh technologies. But what we provide in terms of an ROI proof point, we run our own benchmarks and actually we make them available on our website. Uh so you can see that, for example, for in Q1, and Q1 was a particularly volatile time in the market. There was a lot of movement of all the major indices, and right now our signals are predicting essentially and are helping quants uh trade on those indices. And basically, our signals and this a pure strategy, just using our signal, was able to essentially beat all four indices by anywhere up to seven and a half almost seven and a half percent, right? So there's a big gap into just kind of essentially just following the index and using this intraday predictions that we provide and running a strategy on those. So this is our proof point. This is the ROI that we can provide. So seven and a half percent in one quarter is actually pretty, no bad. Even now, all the financial institutions and all the most of the training is done automatically, uh right, and intra trading, including intra trading. So the question is, I guess, what would be in the head of the some of the investors, I guess, as well. Can you explain why this solution that you're doing right now in Quantum Signal could not exist in the, I don't know, even five years ago? I don't even know ten years ago, but even five years ago? Yeah, yeah. Basically, we're using all the new techniques that have been deployed because of this LLM revolution in the last uh three years, right? So this transformer architecture basically did not exist, right? Did not exist and uh was not in place even four years ago, right? So you had uh some of these big papers that introduced the architecture, and then the LLMs basically started building that. Uh, maybe the papers, so don't quote me on the dates. I think maybe the papers were maybe five or six years ago, and then you know it went into production really in the kind of three-year-ago uh time frame. So um so that's essentially the big uh change. So the technology did not exist, these methods uh did not exist, and that's what what made possible uh the service that we have now and we're kind of offering uh today. And the other interesting thing to note is that the skills to make that possible requires essentially people that are familiar both on with the finance world and the i world. And in the combination of finance and AI is what makes this possible. Otherwise, it's very hard to train models anyway, to train uh AI models and make them work, even though we have the papers. It's very hard for people to make that work. And it's also very hard for AI developers a lot of the time, even though they know how to make the AI piece work, to understand what the finance world needs, right? And in our case, kind of the combination of us working on these hard computational problems for some time, being inside the finance world for uh several years, looking at essentially at the technology gaps over those uh years, over that time frame, that's all of these things kind of created the perfect storm for us to be able to create the solution. So, would you say that this is your um unfair advantage, the combination of the knowledge on AI and LLMs basically that are behind all them and the inside finance world? Because I bet you must have some competitors right now on the market, aren't you? Absolutely. So there are multiple competitors in the market, very few are uh targeting this intraday trading space, and very few are building essentially solutions that are powered by transformers. We see a lot of sort of, if you will, standard or classical ML solutions that uh and other signals that basically that are five-year-old or 10-year-old ML models and ML techniques. That's what we typically see. And then the rest of the competitors are really the uh the things that are built in-house, right? Which are not really, in some sense, they are competitors. Um uh they don't go out to try to distribute them. Really, we don't compete with them uh across basically every every other company we see. Yeah, of course. They have no no, if they develop something good, they they have no benefits to offer it to their own competitors, right? And all H1 trying to, you know, choose the highest returns. Okay, but you mentioned something interesting, which I didn't fully understand, and I think maybe the audience can benefit from explaining it further as well. You said that the transformer AI models are much better than the classical. Is that correct? Yeah, than the classical ML techniques, right? So you have can you explain the difference between sure? Yeah, you have uh I mean 10 or 15 years ago you had all these classical ML techniques, you know, anywhere from uh decision trees and random forests and uh XG boost, like all these different techniques and you know, uh good old regression, right? So so people were using those actively, and actually a lot of the hedge funds, a lot of the quant hedge funds started using those techniques, but markets changed. That's the other big issue here is that you might find a certain pattern in the market, it works for a certain amount of time, and then essentially the edge goes away. You want something that adapts, that can essentially detect new patterns. And the new way of finding patterns are these essentially transformer architectures. That's what the LLMs do. Essentially, they're pattern detectors, right? So they understand that you know what uh comes after, if you give it like essentially half a sentence, it knows what goes after, right? That's why it's able to produce the text, produce the code. Uh, but obviously, instead of using that on text, we want to use it on the market. And in some sense, the problem on the markets is in some sense it's harder, in some sense it's easier, right? So um we in the markets you have this zero-sum games, you have competitive behavior, uh, you have people that are trying to essentially uh mislead uh other competitors, and you also have a lot of noise. So you have to be able to deal with all of that. I mean, that's essentially being a big part of this finance-aware architecture. And um signals. Okay, but this is this is very interesting, and I didn't understand that through the classical, you mean like the old um AI models. But here I'm now surprised because you mentioned that you're still seeing it in some of the uses. So there are still people who are using those old AI models, you know, pring DPT, free open AI, LLM models that we are right now using the Transformers. Because, you know, when you think about it, and I remember that from my studies, that was 15, 20 years ago, they were pretty primitive, right? And very limited in their possibilities. Yeah, you'd be surprised. You'd be surprised still how many people we even find just using uh plain old uh regression models, right? And kind of doing uh things with that. I mean, you know, the the finance world is larger, thousands and thousands of firms, and there are cases where where these models work. I mean, I'm not gonna say that, you know, there are there are essentially effects that can be recognized, uh, but also, yeah, the edge or maybe the return you can get out of some of these detections are kind of becoming smaller and smaller. I think that's why there's more interest now in uh doing something different. Also, I think it's another interesting thing to point out is that people are starting to use LLMs, but not specifically for a trading decision. So LLMs can be good for maybe doing uh portfolio construction. Maybe they can uncover news in the market or a certain stock that might be something might be happening. There might be some report out there that uh essentially mentions a new drug or a new company, a new competitor or something. And a lot of people are now doing research on the market with LLMs. So we see that happening, yes, but that's not training, right? That's kind of research, and also uh a lot of people obviously use LLMs to code new algorithms, so it's easier to now iterate on algorithms and doing essentially different versions of your algorithm and trying different things. So we see all of that happening, right? But again, what we're doing is essentially applying these techniques specifically to build models that are trained on this time series uh data, right? And kind of detecting patterns in in the time series data from the market and then essentially turning that into trades. That's the thing that you know really very, very, very few people can have the capability and are implementing right now. That's that's really great. Okay, again, I'm looking down a little bit of a timeline because it's it's I really enjoying speaking with you. Uh it's all very interesting, but we probably need to wrap up uh shortly. I have a couple of more questions for you about the prediction and the future. Uh, this is my tradition. I I love you know picking the the guest brain on what's gonna happen next. What would be the biggest misconception about the people may have about companies like yours? Okay, well, if you have this technology and it works, why don't you trade it for yourself? Why don't you become a hedge fund yourself? So I feel I have to have to answer that because I think a lot of people are thinking that in the back of their mind. And I think it's it has to do with essentially founder market fit, right? So we came up from essentially math backgrounds, my co-founder Your Thanis and I, math, computer science backgrounds and technology backgrounds, and we kind of had this uh big exposure into the finance world a little bit later. So we're not essentially kind of we didn't grow up in within the finance industry itself, which also kind of gives us this separate perspective, right? So we feel the path for a company is through essentially this VC-backed uh tech uh startup uh path, as opposed to a finance founder that would have essentially maybe tried to build this technology and then started raising capital for investment. That's a different type of investors. You need to find LPs that do that. So the founder fit is definitely different. Uh, the other part is that uh we feel we can disrupt the market and we can make a lot of money in the market by essentially doing this picks and suffils approach, right? So there's two ways to make uh money in a gold rush. You either go for the golden nuggets or you sell the picks and sovels. So we decided to go with the picks and suffels model, and uh so so it's not like we cannot make money or we cannot be successful that way, right? So these are essentially the two key areas where and two maybe misconceptions that might exist or or people might have that in the back of their head. I probably also biased because I spend most of my life in in tech, so we we are all biased one way or another. Okay, and then um I didn't mention that before, I think we didn't spoke about it, but you told me that you have two offices right now, and one is in San Francisco, one is in Paris, you have a small team in Paris. Where is where do you see the strongest ecosystem so far for what you are building? Where is the best market? Yeah, yeah, so we made the decision. So uh Eurdanis is based in Paris, I'm based in San Francisco, so we made the decision as we were starting the company to just essentially everyone stays kind of put and everyone starts building a team around them. Yordanis is is our CTO and and the technical team, he's building the kind of the technical team around him. And also the other good thing about Iordanis is that he's the research director at CNRS, he has a lot of deep connections with a lot of uh PhD students that are coming out of uh the different programs, so we can hire from talents. Exactly, exactly. So that's key, and also for a young company, right? We cannot go out and compete here in San Francisco with AI talent that is here in the San Francisco Bay Area, whereas we can essentially compete and we know people that are essentially coming out of schools and and uh you know other companies. I actually uh I'm not surprised. Definitely it's a very good strategy of being cost effective. Also, the specialist in talent in Europe, I think, is a little bit cheaper than here in the US on the on the local market. And France is very well known right now for having uh excellent uh schools and excellent quantum engineers, so many startups, and I think so much focus as well from the government. Speaking of, I would like to ask you how important it is for you to, you know, work with uh or getting any support from the government, because this is also a very big hot topic right now amongst the startups that are getting a lot of grants or procurement support in in UK, for example. Is there in your space, because you are so much right now commercially driven, is there any support that you may be getting from the governments or not? Yeah, yeah. So and actually we we did get support from the government. So we did get uh two grants, which I'll mention shortly here, and that was another kind of reason, another factor that played into our decision to base the company in France and essentially have a wholly owned subsidiary in the US for the US operations, but the company and the IP and the investments are are made essentially into the French entity, which is the parent entity, and we felt that we were already connected through your nice with a lot of the the French entities and and the French organizations that can provide early support. And initially we needed that, right? So now, yes, we're getting into the situation where we are essentially building the product and going out and getting customers and having uh some revenue, and we're trying to grow that. Uh, so we're not looking as much towards uh grants, but initially, two years ago when we were just getting started, basically we did apply for grants and we did receive the Innovap Leader um award from Paris. And then we also received another award from uh yeah, from de France region, which was basically called Pac Quantique. It um it uh talks, uh it's it's it's awarded to new young quantum computing startups that are building innovative quantum computing applications, and they're working with uh with a partner, right? And uh the partner, and this is public, the partner is for that award is Societe Generale. So uh we kind of got together and applied and won that award. And so these are uh two of the key awards that we won early on and helped us uh significantly as we were kind of going through the first stages. That's great. So you basically proving that those early stage support um in grants, in cash, in contracts from government is really important. If you would not set up the company in France, would you have the chance to get a similar support from US government or not really? I know there are a lot of grants that exist in the US as well. I think in some cases it's tricure. There is a lot of awards that go directly to academics. This is the case obviously in Europe as well, maybe to a lesser extent. Yeah. Um, I don't want to say that it's harder because, in some sense, even though I'm based here in San Francisco and I know the the US side of the equation, I actually know it less. Now that we were based in France, we went after some of these awards. I I know how the system in France now works better. And it helps us quite a bit. And maybe maybe San Francisco is also more competitive, or I don't know, from around the world. It could be. Yeah, yeah, absolutely, absolutely. I'm not discounting that it. And you mentioned Obusit that you have this, you basically were enabled from their day first to the French ecosystem because of your co-founder and his involvement in CNRS. I I know, but maybe you can explain what CNNRS is actually, because I don't think our audience is aware. Yeah, yeah. So it's one of the uh biggest uh research institutions in France. Yeah, so they have obviously uh multiple departments uh in different areas. Jordanis was leading essentially quantum computing, the research director for uh for quantum computing uh at CNRS. Um yeah, not sure what else to say. This is uh this is uh yeah, one of uh it's very well recognized. I think all of Europe now is in all of Europe, it's very well recognized and produces uh a lot of good research, a lot of essentially good uh researchers are coming out of that. And yeah, it it helped us uh significantly. Well, having a Urtanis there, basically, we played our our strong card, and that's why we built the company like we did in France. Yeah, I think it's a perfect setup. Uh I think it's a perfect setup to be on both sides and get the best from both worlds. Um I would like to ask you right now, what kind of milestone should we be watching for quantum signals in the next 12 to 24 months? Yeah, so we are focused on two things. One is producing uh more and better signals, right? So we are publishing our benchmarks on uh on a monthly and quarterly basis. Uh so we will be improving uh the quality of the signals and the amounts of different signals and these intraday predictors that we are producing. We'll be doing futures, the futures US futures contracts, as well as single name equities in the US, so single name stocks in the US. So that's coming in the next few months, effectively. More customers. So we are looking forward to making some public announcements on customer acquisition and then customers using our product actively for trading bigger and bigger portfolios. Yes, and then also raising our seed round. So we would be essentially at some point be making an announcement on how the seed round is going and size, investors, that that sort of thing. How many clients would you like to acquire in the next 12, 18 months? So um my goal is to finish uh 2026 with uh somewhere in the 15 to 20 uh range. Well, uh, I think yeah, I think we can maybe sign one or two every month for the next uh few months until the end of the year and kind of get into that 15, maybe you know, uh 15-20 range. That sounds like a lot of work. I will be very busy for the next couple of months. And uh, if I can ask you, I I just shouldn't because really of the time, but I can stop myself. How do how do how are you winning those clients? Yeah, so the good thing is that you know when when you're publishing results and predictions on the market, it's actually, you know, the proof is there, right? So you can't hide from it. So people are looking at the results, previous results, previous performance. They are signing up to basically start tracking real-time performance. So it's very kind of black and white, right? You can't, it's not uh qualitative at all. It's all in the numbers, and basically every day we're producing uh these predictions that basically say uh whether a certain index, for example, the SP 500, uh, if it's gonna go up, if the market is moving one way or another, and so people can trade on this, right? Um, so that's the core of our product, and so it's very basically we get tested on a daily basis, right? And in most cases, people want to see it to see a little bit more, collect a few more samples on a daily basis, and they can make a decision. So that's that's pretty simple actually for us. I guess the hedge funds they can be very fast on making those decisions, right? But how about the the bonds? Uh, do you have any secret? How are you escaping those famous 12 months enterprise sales cycle or or not? Yeah, unfortunately, I still have not found the secret for that. If you find it or if you want other guests finds it, please let me know. But yeah, it's it's the same kind of uh very slow cycle. But yeah, what what helps is to essentially have the support of the business, right? So that's always the case. So if you have the support of the business, then you can push from both directions. So the business side can can push internally, you can push the procurement externally. It still takes essentially forever to sign some of these contracts. The the contract we did sign with uh the European Bank was definitely very painful. It took uh months upon months. Yeah, I don't have any, and but you're right, the hedge funds are much faster. So we're kind of planning on maybe adding another bank within the next before the end of the year, uh if we can kind of squeeze it in there uh with uh the procurement cycle. But hedge funds we expect that we can be add maybe Yeah, they are like okay, great, perfect. Last question, really last one. That would be about the predictions. So, Yanni, what would be the boldest claim that you are willing to make? You can make it about quantum industry, you can make it about your finance industry, you can make it about your company. Yeah, free shot here. Yeah, yeah, I think yeah, I I'm gonna concentrate on kind of what we do and what we know best, which is essentially uh building these AI agents. I think they're gonna take off very, very uh quickly. And I think we're even gonna see some prosumers or essentially individual quants using them. I think that's essentially the the new way that people are going to be trading. No more essentially discretionary trading, or it's gonna become less and less of a of a thing. And essentially we're gonna have we're gonna be seeing more and more AI agents basically actively trading in the market. Yeah, we're gonna see this back and forth we see from with LLMs right now. So you have LLMs and then you have user A prompting an LLM, getting essentially uh a certain response on that prompt, and then user B is prompting the LLM in a different way, right? And a different prompt and getting essentially a different response. This is the kind of thing we're gonna start seeing with uh AI uh trading strategies. You're gonna have one quant or one trader basically prompting uh these AI models uh in one way and getting one type of strategy, one type of trading strategy out of the model, and then another way essentially prompting another user prompting it in a different way, getting a different type of strategy out of that. So that's kind of trans fiction. The agent will be fighting with each other. Yeah, yeah, yeah, it will it will happen for sure. It will happen. I think that's the only uh that's uh the only uh certainty that I see in the trading future. That sounds very exciting, but in the same time very terrifying. It is a little bit, yeah, yeah. It's not clear exactly. I mean, we've seen some in the past like algos kind of going uh uh haywire a little bit, especially high frequency algorithms, caused some flash crashes as they were known. Uh so we will see some some of the effects of this over the next few years for sure. But I think you know, we need to prepare for that. Uh, you know, there needs to be regulatory framework, other types of oversights, other types of uh um other types of uh guardrails there for that. But I think you can't you can't stop the progress of technology. I can assume that before the regulation will pop up, a couple of people will get very, very rich. And that's always happened, right? That was the case. But uh, I wanted to ask you, what's the time frame? Um, are we talking about this future coming into five years, ten years? No, I think um I think the technology is coming much faster. We've seen from the uh advancements in AI and how um the advancements in LLM and what we can do now with coding that um um you know that things are moving very, very fast. We're talking about uh frameworks or and time frames of months, you know, maybe 12 months. Let me ask you a a different question. In what time do you think most hedge funds will have either solution like quantum signals or something similar? So basically agent trading, intraday trading, but also different kind of tradings that is in the months, you think, or are that gonna happen a year, two, three? Yeah, I think I think we're talking about maybe a year and a half to uh two, three years, I think. Wow, I mean we've seen what has happened with uh yeah, we've seen what has happened with LLMs and you know coding and all of that. I think a lot of the financial industry is conservative, right? Uh they're already using LLMs, so they already see the power, and so now giving them something that basically transfers that same type of power, same type of tooling, but transferred basically to their own uh domain, I think they're gonna start making the switch very, very quickly, right? So, yeah, perfect. And that's what I wish for you for quantum signals. And I hope if we met in the two years from now, you would be like, I don't know, raising C maybe B or C, or maybe preparing for IPO, who knows? Seems like a lot of opportunities, and you will be definitely very busy. So I hope you will find time to come back at some point. Of course, yeah, yeah. Happy, happy to do so. Yeah, yeah. Well we'll uh yeah. Thank you. It was a pleasure. Great discussion. Thank you so much, Eva. Thanks for the invitation and for the great discussion. Thank you. Bye bye. Take care, take care. That was Blonde and Quantum. Thank you for joining me on this journey through the quantum business frontier. If you like the episode, please review another Spotify or Apple Podcast and help more people discover the quantum world without needing to untangle the theoretical physics. See you next time. Unless the cat changed the time on first.