Listen to our latest episode featuring Growi Finance and learn how a research-driven, mid-frequency mean-reversion strategy became one of the top-performing vaults on Hyperliquid, and why risk-by-design matters more than chasing yield.
In this episode, Nertila from Trading Strategy discusses with Valentin and Miguel from Growi Finance. Growi Finance is a quantitative asset-management firm building on-chain yield products to institutional standards. It describes itself as the allocation layer for digital asset yield, serving both retail investors and professional allocators.
Play video: Episode #14: Growi Finance- View Growi.fi Vaults and Yields
- Watch the Growi.fi episode on Youtube
- Listen to the Growi.fi episode on Spotify
Any questions? Join our community Discord server.
Transcript:
Welcome to TradingStrategy.ai, the podcast where you can learn more about the essentials of DeFi trading in the new world. This is your go-to show for automated trading strategies and market insights. Welcome to the conversation, and this is your host, Nertila.
Nertila: Hello, everybody. Welcome to today's podcast at TradingStrategy.ai. Today we have two guests from the Growi Finance vault, Miguel and Valentin. Thank you so much for joining me today.
Valentin: Thank you.
Nertila: We are very curious about your backgrounds. When did you start working in the crypto industry, and how did you end up in the vault business?
Miguel: Our origin is more academic than most vault stories. Growi started in 2020 as a research collaboration between Valentin, our CEO, and Gorka, another Growi founder. Valentin is an aerospace engineer with 15 years of modeling at Airbus and a master's in international finance, and he has also spent years teaching applied mathematics at university. Gorka, then a student, proposed building a quantitative strategy for digital assets as his final-year thesis.
As for me, I have always worked in business development, including at several startups unrelated to crypto. I've been in crypto since 2020. My most recent role was at a publicly traded company, managing major accounts, including Fortune 100 and IBEX 35 accounts, in the private security sector.
That thesis, started in 2020, turned into four years of research. During those years, Gorka ran the models as a prop trader, so the strategy was being validated with real capital on centralized venues long before it ever touched a vault. By the time Hyperliquid vaults appeared, we weren't a team looking for a strategy. We were a strategy looking for a distribution rail, and Hyperliquid vaults solved our hardest problem: how does a small Spanish quant team give anyone in the world transparent access to an institutional-style strategy without building fund structures first? We went live on September 1, 2024, and everything has been real since that vault went live.
Valentin: I'm an aeronautical engineer. I've been working at Airbus, the aircraft company, for about 19 years. In parallel, I worked in the Applied Mathematics Department at the Aeronautical School, where I supervised master's and bachelor's theses related to financial mathematics. I've even written a book, a popular-science book about financial mathematics and crypto, specifically Bitcoin. Then Gorka, as a student, suggested developing a strategy, and that's where our journey started.
Nertila: That's very interesting. So tell us about Growi. What is it about? So far, it has been the number one vault for several weeks.
Valentin: Growi Finance is a fairly young company; we've been around for a couple of years now. Initially we focused only on Hyperliquid, and now we've also moved into creating other products. What we try to do is get alpha for the people. We always try to apply our math knowledge to get an edge. For example, in lending we now earn around one to one and a half percentage points more than our competitors. On Hyperliquid, we have a very good recent track record, and we now have the biggest AUM among the top vaults. We try to bring this alpha to people in every product we develop. So far, Hyperliquid is probably the most consolidated one; as you say, we're in the top two there.
Nertila: Are you running multiple strategies or a single strategy?
Valentin: On Hyperliquid, we run a single strategy, but that single strategy is made up of multiple bots. It's the same strategy principle, the same mathematical equations, but each position is managed by many bots. For example, our BTC position isn't managed by one single robot. We typically have around 80 bots managing that position, and what differs between them is that they work on different timeframes. The timeframes are similar, because we work in what's called mid-frequency, in the range of minutes. It's not high frequency, which works in seconds or milliseconds. Our bots work on timeframes between roughly five minutes and one hour, all running the same strategy, each on a different timeframe.
Nertila: Are you running a specific kind of trade, or do you try to do a bit of everything?
Valentin: We are focused on mean reversion. For the moment, it's one single mean-reversion strategy, and that's it.
Nertila: You started very early on Hyperliquid, when there was no historical data. How did you deal with that, considering you both come from mathematical backgrounds and work with data and numbers?
Miguel: The mathematical model wasn't developed on Hyperliquid data. It was researched and traded on venues that already had depth. Mean reversion is a property of market microstructure, not of a specific exchange. If anything, a young venue with retail-heavy flow and frequent dislocations is a more favorable environment for it. What we did treat as a genuine risk was the venue itself: a new DEX, new infrastructure, new platform risk. There, we did what any risk-conscious team does. We started with our own capital, sized conservatively, and scaled as the venue proved itself and Hyperliquid became the category leader. It was a sized bet, not a leap of faith.
Nertila: Let's look at some historical events on Hyperliquid. We can see from the Hyperliquid PnL curve that there have been two incidents where the profit and loss curve dropped more than 70%. I'm not sure whether depositors panicked, so I'm curious what their reaction was. Then, from that drop, it comes right back up. What happened there?
Valentin: First of all, PnL is not a good indicator of risk, because PnL is in dollars, and the dollar amount depends on how much capital you have at that point. To assess a drop, you need to look at the TWR, the time-weighted return. If you go to Hyperliquid and look at the strategy's maximum drawdown, it says 12%. But that 12% isn't representative either, because Hyperliquid doesn't record every point of the history. For older NAV points, they record roughly one point every couple of weeks. Our real maximum drawdown, measured on daily NAV as is standard in the industry, is 20%. So 20% is much more realistic than the 70% you see on the PnL chart.
I understand Hyperliquid's choice: PnL is very direct and easy to calculate, but it's probably not the best variable to show. And their 12% maximum drawdown isn't realistic either. So in reality, it's very difficult to know a strategy's drawdown on Hyperliquid. That's why there's space for people like you, who can properly record a strategy's metrics, such as maximum drawdown or annualized yield.
As for the drawdowns themselves, they were difficult, of course. There were two episodes that affected the whole industry, including us: October 10 and February. It was hard, but our strategy managed to overcome them because it is sized for this type of event. We put more effort into being robust than into getting more yield. What matters most to us is a high Sharpe ratio. We take all the risk parameters into account, such as the Calmar ratio, and we try to maximize the risk-reward ratio. That's why we're able to overcome these extreme events: we design for them.
We also run stress tests, subjecting the strategy to extreme scenarios that aren't even realistic, like BTC going to zero or BTC going to one million, so that it can pass when real difficulties arise. The February event was super hard. It was a five-sigma move, something very extreme in the market. It was difficult, but as I said, we are sized to withstand that kind of stress event.
Nertila: You mentioned risk. How do you manage it?
Valentin: We manage risk from the design stage, and then also in real time. From the design stage, as we say, the hardest thing is not getting yield; it's not getting liquidated. It was a pity, because I know how hard people worked to get there, but on October 10 many of our colleagues were unfortunately liquidated.
So from the beginning, risk is on our minds when we design a strategy. We run stress tests and we generate synthetic paths to test how robust the strategy is. We're very sophisticated with these synthetic paths; we can reproduce paths very similar to real ones, accounting for volatility clustering and similar effects. Then we run specific stress tests on top of that. In real time we also have some safeguards, but because we are so strict in the design phase, so far we have never had to take the critical measures we would take if we got close to liquidation.
Nertila: How do you choose what to trade? Some of the tokens vaults trade are quite illiquid, some have small caps, and there have been episodes of them going down significantly.
Valentin: We try to trade the most liquid ones. Our trading universe has two requirements. The first is historical data, because our strategy is statistical, so we need enough data to test. We require data going back to 2021, so we can also analyze downtrends. That means we don't trade the most recent tokens, although we may make some exceptions; we're currently considering whether to include HYPE. The second requirement is that the token allows leverage higher than 10x, which usually means the venue considers it liquid enough, since liquidity is one of the factors in how leverage is assigned.
Right now, I think only one of our tokens has changed status and has 5x leverage, which is INJ. But in general, what we trade are the most liquid tokens on Hyperliquid. I think we have 19, which are more or less the 19 most liquid tokens. From time to time we need to update the list, and we will before the end of the year, because there are new tokens on Hyperliquid. But in general, they are the most liquid ones.
Nertila: Growi Finance became a top vault, but were there times you doubted yourselves working in the crypto ecosystem? What are some of the most painful or valuable lessons you've learned along the way?
Miguel: Of course. February 2026 was a difficult moment. Watching a 19% drawdown live, with institutional capital on board for the first time, tests your conviction in a way no backtest does. The lesson: study your drawdown profile before it happens. We knew what normal "bad" looked like, so we didn't touch the model. Discipline means deciding in advance what would make you intervene, so you never decide under stress.
Also, in our case, distribution is harder than generating alpha. Four years to build a strategy, and it still turned out to be the easy part. A verifiable 100% track record doesn't sell itself.
And we have to talk about October 2025 as the flip side. That episode scared the entire crypto ecosystem, and for us it was our best month ever. The lesson we learned is to know precisely which regime fits your strategy, because that's what lets you keep your systems running while everyone else is pulling the plug.
Valentin: To complement what Miguel said: as a company working with Hyperliquid and its ecosystem, it's very hard. On one side, Hyperliquid gives you a lot of eyeballs, which is why we can be there. On the other side, it's very difficult to run a business on Hyperliquid, because vaults are very strict. There's the rule that the manager has to hold 5% of the vault, which means you cannot scale; you can only manage a certain amount of money and no more. And you only get paid when clients withdraw their money. I understand why it was designed that way at the beginning, and now they even want to phase out vaults. But it's very hard for managers, because if you have loyal customers who love your product, you never get paid. As a company, that's horrible. So on one side, it's great to be on Hyperliquid; it's a very good venue. On the other side, for vault managers it's very difficult.
I've communicated this to the Hyperliquid team. I think vaults are a very good product. I know they want us to move to HyperEVM, and we are going to HyperEVM with a smart contract as well. But the vault is something special, because people don't need to trust your smart contract. Some people are afraid of all that, or don't even know how it works. They know how Hyperliquid works, and that's it. For those people, the Hyperliquid vault is a very nice solution. I can even imagine mid-tier traditional asset managers coming to manage vaults on Hyperliquid. But that's our opinion; they clearly have another one, because they want to phase vaults out. So it's very difficult and complex.
And what Miguel said is also very surprising. Last year we generated more than 40% yield, which is extremely competitive in finance. Even though we've had some success, you would expect even more success with that yield. That's been difficult.
Nertila: Where do you think capital and risk will accumulate in the future?
Miguel: Our thesis is that capital accumulates where trust is cheapest to verify, and that's why vaults and curated lending are growing so fast. On-chain asset management gives you real-time, verifiable positions and performance, and a level of auditability that traditional fund structures can't match. We think the allocation layer, connecting strategies to both retail and institutional capital with proper risk curation, is where the next wave of value settles.
Nertila: This will be the last question. We see you are expanding to other perp DEXs and to lending. What is next for Growi Finance? What will Growi Finance 2.0 look like?
Valentin: Since the beginning, we've wanted to bring the crypto ecosystem to users. When we started in crypto, what we found was amazing: you are the custodian of your own tokens, there's a lot of disintermediation, and there are many opportunities. Morpho and other vaults, for example, are something unique. You can interact without any intermediary or bank and earn yield yourself, yield that is usually hidden by the banks.
So there are a lot of opportunities, but each product is very complex. Even lending vaults are complex, because you need to consider the collateral, the oracle, the audits. We've recently seen some issues in lending, which is the simplest product; what we do with perpetuals and leverage is quite complex. But even the simpler products are complex. So what we want is to give yield to ordinary people and bring people from traditional finance into the DeFi ecosystem, and for that, we want to make it as simple as possible.
We are professionals; we know risk and can assess it. Our idea is to integrate both our own products and third-party products into the Growi protocol. A person deposits into our vault, our protocol, and we distribute that capital across our own products and different third-party protocols, so they get the best yield in real time without needing to understand how Aave, Morpho, or Pendle work. Our vision is what we call the allocation layer, so people can participate in the whole crypto ecosystem by investing with us. Whether we'll manage it or not, we'll see, but that's our vision.
We also didn't mention that we already work with institutional clients, some of them very large. One of them manages $30 billion, and we have strategies running with them. Some of our institutional clients are crypto-native, but others, like the one with $30 billion, come from traditional finance. Right now it's mostly professionals and institutions from traditional finance coming to crypto. I would like to bring everyday people to crypto too. That would be my dream, so they can benefit from this great ecosystem and everything it offers today.
Nertila: Miguel and Valentin, thank you so much. The conversation was very insightful, and I really appreciate you sharing your ideas and opinions about crypto. I hope to have you on the podcast again.