Martin Schmid, Matej Moravcik and Rudolf Kadlec have turned an old DeepMind problem - how to train an agent to win when it cannot see the whole game - into a higher-stakes target: public markets. EquiLibre Technologies, the Prague AI trading lab the three former DeepMind researchers co-founded, is now valued at more than $500 million after an undisclosed Series A led by Creandum, TechCrunch reported on June 30.
The amount of the Series A was not disclosed. That matters, because the valuation is doing most of the signaling here. Creandum vice president Cameron Sellers told TechCrunch the deal was the largest single investment the firm has made "in one go into a company."
The founders' pitch is not that they discovered automation in markets. Quant funds have spent decades turning statistics, infrastructure and latency into an edge. EquiLibre's bet is narrower and more technical: the same reinforcement learning and game-theoretic machinery that helped its founders build poker systems can be adapted to markets where competitors adapt, information is incomplete and the reward function is brutally legible. As Schmid put it to TechCrunch, the score in trading is simple: "how much money did the agent make?"
The founders came from games, not finance
Schmid is EquiLibre's CEO, Moravcik is its chief science officer and Kadlec is its CTO. EquiLibre's own site describes Schmid and Moravcik as Ph.D.s in algorithmic game theory and former DeepMind and IBM researchers; Kadlec is described as a Ph.D. in episodic memory modeling, also formerly at DeepMind and IBM. Schmid and Moravcik co-authored DeepStack, while all three are listed by EquiLibre as co-authors of Player of Games.
That background is the center of the company, not a credential line. DeepStack, published in 2017, showed that an AI system could defeat professional players in heads-up no-limit Texas hold'em, a poker variant where the player does not know the opponent's private cards and must reason under uncertainty. The founders later worked on Player of Games, a more general algorithm combining search, self-play learning and game-theoretic reasoning across perfect and imperfect information games.
EquiLibre's move into trading follows directly from that research history. The company says on its homepage that its systems use reinforcement learning agents that "learn, adapt, and compete" and that its models trade billions of dollars every day. Those are EquiLibre's claims, not audited operating metrics. TechCrunch separately reported that, through a partnership with Tower Research Capital, EquiLibre's algorithms have traded billions in daily volume across the S&P 500 and Nasdaq.
Schmid is explicit that the draw was the build, not Wall Street as an institution. He told TechCrunch he is not motivated by making markets efficient; he is motivated by "building new things that have never been built before." That distinction helps explain why EquiLibre presents itself as a research lab first and a finance company second. It also explains why venture investors are willing to underwrite a company whose most important outputs may look less like SaaS revenue and more like model performance, execution quality and negotiated economics with trading partners.
The valuation is a bet on direct monetization
EquiLibre's reported valuation is easier to understand if the company is viewed less like an enterprise AI vendor and more like an AI research group with a path to cash conversion. A better trading agent does not need a long enterprise sales cycle to prove its value. If it works in live markets, the feedback loop is immediate; if it fails, the market also says so quickly.
That speed is why the missing details matter. EquiLibre has not disclosed the Series A size, total funding raised or the exact economics of its Tower Research Capital relationship. TechCrunch reported that the startup claims its agents have recorded "zero negative months since inception," first in crypto markets in 2025 and then on stock exchanges. That is a performance claim from the company, and it should be read as such until the underlying track record, capital base, risk limits and fee structure are visible.
For founders, the more interesting lesson is where EquiLibre chose to be exceptional. The company does not market itself as a tool for retail traders or as another dashboard wrapped around market data. Its public site says the company has "no sales" and "no marketing" and judges itself by the quality of its models and algorithms. That is a posture only a narrow set of AI companies can take: a hard technical team, a valuable domain, and enough investor belief to let research compound before a conventional go-to-market motion.
Prague is part of the strategy
EquiLibre is also a geography story. The founders built the company in Prague after working at DeepMind and IBM.
That is not just lifestyle positioning. EquiLibre's advisory board, as listed on its website, includes reinforcement learning and game AI figures Rich Sutton, Michael Bowling, Csaba Szepesvari, Michal Pechoucek and Murray Campbell. The company is trying to use Prague as a base for deep technical work while selling into one of the most competitive markets in the world.
EquiLibre's advantage, if it has one, will not be that reinforcement learning is unknown to finance. It will be whether Schmid, Moravcik and Kadlec can keep translating their imperfect-information research into agents that survive changing markets, transaction costs and competitors that learn back. Poker gave them a benchmark. The market gives them a scoreboard.