From Poker to Profits: How DeepMind Alumni are Revolutionizing Trading

Former DeepMind researchers are pivoting from defeating professional poker players to managing billions in market volume. Their Prague-based startup, EquiLibre Technologies, has achieved a staggering $500 million valuation by applying reinforcement learning to the high-stakes world of quantitative finance.

Translating Poker Strategy to Wall Street

The core innovation driving EquiLibre is the transition of reinforcement learning (RL) from games of imperfect information to the complexities of the stock market. The founding trio—CEO Martin Schmid, CTO Rudolf Kadlec, and CSO Matej Moravcik—previously gained fame for developing DeepStack, the first AI to defeat professional no-limit Texas hold ’em players.

The logic is a natural evolution: both poker and trading involve making optimal decisions under uncertainty with clear, measurable outcomes. As Schmid notes, the "scoring" in trading is remarkably simple—the ultimate reward is capital gain. By utilizing RL, where models learn through incentivized feedback loops, EquiLibre has moved beyond gaming to execute trades across the S&P 500 and Nasdaq.

Massive Scale and Proven Performance

EquiLibre is not merely running simulations; it is actively participating in global markets. In partnership with the quantitative firm Tower Research Capital, the startup’s algorithms have been managing billions of dollars in daily trading volume.

The startup's track record is particularly notable for its consistency. After an initial rollout in the crypto markets in 2025, the company expanded into traditional equities, claiming a "perfect record of zero negative months since inception." This level of stability is a major draw for venture capitalists like Creandum, which recently led a Series A round that marked the firm's largest single investment to date.

The Race for Compute and Talent

While EquiLibre has successfully scaled to a $500 million valuation, it faces intense competition from established trading giants like Jane Street, which utilizes tens of thousands of high-end GPUs and combines RL with Large Language Models (LLMs).

To compete, EquiLibre is focusing on a "lab-first" approach rather than a traditional finance mindset. Their strategy involves two key pillars:

  • Efficiency over Brute Force: Rather than relying on massive GPU clusters, the team aims to "get more from less," optimizing algorithms to squeeze higher performance out of limited compute.
  • Strategic Infrastructure: The company plans to build one of the largest compute clusters in Central and Eastern Europe (CEE) to scale its research capabilities.

By basing themselves in Prague, the founders have also tapped into a specialized Czech diaspora from companies like Google, allowing them to build a high-caliber team of 25 experts outside the hyper-competitive San Francisco ecosystem.

Key Takeaways

  • Algorithmic Evolution: EquiLibre is successfully porting reinforcement learning techniques used in professional poker (DeepStack) to manage billions in daily S&P 500 and Nasdaq volume.
  • Explosive Valuation: Following a successful Series A led by Creandum, the startup has reached a $500 million valuation, fueled by a reported "zero negative months" track record.
  • Efficiency as a Moat: Facing giants with massive hardware advantages, EquiLibre is focusing on algorithmic efficiency and building significant compute infrastructure in the CEE region.