Ex-DeepMind researchers turn poker AI tech into $500M trading powerhouse

Three researchers who previously developed a groundbreaking AI system to compete against world-class poker players have now pivoted their expertise toward financial markets. Their startup, EquiLibre Technologies, has reached a $500 million valuation following a Series A funding round led by Creandum—marking the venture capital firm's largest single investment to date.

From game theory to trading algorithms#

The shift from poker to quantitative trading may seem unconventional, but both domains share a critical characteristic: they reward agents that excel under conditions of imperfect information and uncertainty. Reinforcement learning, the machine learning approach underlying the original poker breakthrough, allows algorithms to optimize behavior by learning from rewards. In trading, Martin Schmid, EquiLibre's CEO, notes that the feedback signal is straightforward—simply whether the algorithm generated profit.

Partnered with hedge fund Tower Research Capital, EquiLibre's models have been executing billions in daily trades across major U.S. stock indices and cryptocurrency markets since launching in 2025. The team claims a spotless track record: no months with negative returns since inception.

The founding team's journey#

Schmid and his co-founders—CTO Rudolf Kadlec and Chief Strategy Officer Matej Moravcik—built their original poker-playing system, DeepStack, while serving as visiting PhD students at DeepMind's research office in Edmonton, Alberta. That program became the first AI to defeat professional no-limit Texas hold'em players. When the Edmonton office closed in 2023, the trio chose to relocate to their native Czechia rather than remain in Silicon Valley.

The decision proved strategic. By returning to Prague, they could assemble a team from the local Czech diaspora already embedded at tech giants, building a 25-person operation that has remained stable even as competitive AI ventures proliferate elsewhere. Schmid notes that the lack of constant disruption from newly funded AI startups makes retaining talent easier there than in San Francisco.

Reinforcement learning's moment#

The startup's momentum reflects a broader shift in how the industry views reinforcement learning. When EquiLibre began four years ago, Schmid recalls skepticism around the approach; today it has become standard practice in trading. The company gained particular attention as RL gained credibility for financial applications.

That credibility hasn't gone unnoticed by established competitors. Trading powerhouse Jane Street operates tens of thousands of high-end GPUs and explicitly leverages RL alongside large language models in its trading operations. EquiLibre, by contrast, is pursuing an efficiency-focused strategy—aiming to extract greater performance from a smaller compute footprint while planning to build one of Central and Eastern Europe's largest compute clusters.