The DeepMind trio who built a poker AI are now making money for quant hedge funds

None of the three founders of EquiLibre Technologies had ever worked in finance before creating an algorithm that now moves billions in the markets. Yet, from 2025 to today, their AI has never closed a single month in the red: a record of zero negative months, just as CEO Martin Schmid says. The widespread belief is that to win on Wall Street you need networks, insider information, and a wolf's mind. Instead, EquiLibre comes from a completely different universe — academic research and online poker games. The game-changing theory is this: the skills needed to beat the big markets are no longer those of the financial sharks, but those of scientists capable of teaching a machine how to learn on its own. It's called reinforcement learning: systems that gain experience, make mistakes, are rewarded, and adapt — exactly like human intelligence, but at a frightening speed. Who are these three? Martin Schmid, Rudolf Kadlec, and Matej Moravcik: all three were visiting PhDs in the DeepMind office in Edmonton, Canada, when they created DeepStack, the first AI capable of beating no-limit poker champions in Texas hold'em. None of them wanted a career in finance, at least not at first: Schmid says it clearly, “I'm not doing it to make the markets more efficient. I do it because we enjoy building things that don't exist yet." After DeepMind, they returned to Prague with a group of friends and colleagues — many of whom were part of Google's Czech diaspora — and founded EquiLibre. They decide to stay there, far from Silicon Valley, because – in Schmid's words – "it's easier to keep talent here: there's not a new sexy startup every two months trying to snatch them away from you." Today there are only 25 people, but they manage algorithms that, in partnership with Tower Research Capital, trade on the S&P 500 and Nasdaq, with staggering daily volumes. The story of EquiLibre is also a case of timing: when they started, reinforcement learning was viewed with skepticism — now it's the standard, so much so that even Jane Street, one of the world's giants in quantitative trading, claims to use it along with advanced language models and tens of thousands of GPUs. But EquiLibre, with much more limited resources, focuses entirely on efficiency: "We have to do more with less," says Schmid. The paradox? In an industry where automation should lead to a few winners and many losers, Schmid sees another possibility: "This is not a game where the winner takes it all." And the surprising detail is that EquiLibre doesn't even call itself a financial company: it sees itself as a research laboratory that, by chance, has found a way to print money with an algorithm. Try to imagine: three former AI researchers, who have returned home almost out of nostalgia, start playing with the world markets — and within four years they are worth more than half a billion dollars. But the real point is that today, engineering creativity counts more than financial cunning. If you think the future of Wall Street is written by former Goldman Sachs traders, perhaps it's time to reconsider your bet. From now on, it could be the reinforcement learning nerds who decide who wins, not the sharks in suits and ties. If this story of science and algorithms has opened a new window for you, you can mark I'm In on Lara Notes — it's not a like, it's a way of saying: this idea is now mine. And if tomorrow you find yourself telling it to someone — maybe in front of a deck of cards, or a stock market monitor — on Lara Notes you can tag those who were there with Shared Offline, so that conversation stays alive. All this coming from TechCrunch, with 1 minute saved.