EquiLibre's $500M Prague Round Turned 25 People Into a Sourcing Map
Creandum's largest-ever check just legitimized self-play RL in trading. Here is how to source the tiny, mostly Central European talent pool before Jane Street does.
On June 30, Creandum led a Series A into Prague-based EquiLibre Technologies at a valuation north of €438M ($500M), the largest single investment in the firm's history. The company has 25 people. That works out to roughly $20M per employee, and it just put a price tag on a talent pool most sourcers have never built a search for: engineers who ship reinforcement learning into live financial markets.
If you run engineering hiring at a quant fund, an AI trading startup, or any deep-tech company that competes for the same profiles, the round is more useful as a map than as news. It names the feeder employers, the geography, and the specific flavor of RL that matters. This post walks through how to read that map.
The round in one paragraph
EquiLibre was founded by Martin Schmid, Rudolf Kadlec, and Matej Moravčík, the three former DeepMind researchers behind DeepStack, the first AI to beat pros at no-limit poker. They incorporated in 2022, took a pre-seed from Credo Ventures (early on ElevenLabs and UiPath), then a $10M seed led by Blossom Capital at a €122.8M valuation, then this Series A from Creandum. Turing Award winner Rich Sutton is both an advisor and an investor. Their algorithms, run in partnership with Tower Research Capital, already trade billions in daily volume across the S&P 500, Nasdaq, and crypto. Part of the round is earmarked for hiring more deep learning researchers and engineers in Prague, plus what they expect will be one of the largest compute clusters in Central and Eastern Europe.
That $20M-per-head number is the frame. It tells you the market is pricing a specific, narrow skill combination, and it tells you where the offer bar sits when you go head-to-head.
"RL engineer" is the wrong Boolean
Most sourcers who read the headline will paste "reinforcement learning" AND ("quant" OR "trading") into LinkedIn and call it a day. That search returns garbage because the skill EquiLibre is actually paying for is not generic RL. It is self-play reinforcement learning under imperfect information, the DeepStack lineage. Training an agent by letting it play itself millions of times in adversarial conditions is not how most ML in trading has historically worked. Most quant ML is still supervised, fitted to historical data. That is why the intersection is tiny.
If you are serious about reinforcement learning engineer hiring in this niche, source from the communities that actually work on the underlying math:
- Counterfactual regret minimization (CFR) authors and their citation graphs
- Game-theory equilibrium and imperfect-information research groups
- Poker-solving projects (DeepStack, Libratus, Pluribus, ReBeL)
- NeurIPS RL workshops and ICAIF (AI in Finance) accepted papers
- The ML Prague conference speaker and attendee lists
None of those are LinkedIn-native signals. They live in arXiv, GitHub, conference PDFs, and personal sites. This is the friction that keeps most recruiters stuck on keyword search, and it is the specific reason we built Refolk: you describe the person in plain English ("published on CFR or self-play RL, currently at a quant fund or ML research lab, open to relocating to or already in Central Europe") and get a ranked shortlist that stitches GitHub, LinkedIn, and open-web signals together.
The feeder-employer list is unusually short
EquiLibre publicly names its team as drawn from DeepMind, Google, Jane Street, G-Research, and Optiver. Add Tower Research Capital (their trading partner) and the adjacencies most sourcers should already be watching: Jump Trading, Two Sigma, Citadel Securities, Renaissance Technologies. That is roughly nine to ten companies. For most engineering searches that would be a starting set. For this one, it is close to the entire credible universe.
Concentration is a gift. It means outbound for quant AI trading recruiting can be nearly exhaustive within a quarter, which is not true of general ML hiring. Build the list once, keep it current, and you will see the same 400 to 800 people move across it for years.
Where the pool actually lives
The other half of the map is geography. Refolk's internal data on professional-network signals shows that the intersection of "reinforcement learning" plus "quantitative trading" across Czechia, Poland, Germany, and the UK returns only single-digit exact matches. The emergent hubs are London, Kraków, Munich, and now Prague, with Jump Trading, ING, and JERA Global Markets showing up as current employers alongside the obvious names.
Central European sourcing has its own rules. LinkedIn coverage is thinner. GitHub coverage is often better than LinkedIn. Personal sites and university group pages (Charles University, CTU Prague, TU Munich, University of Warsaw, ETH Zurich) matter more than in the US. Prague deep tech talent tends to be findable through the prg.ai network, ML Prague, and the CTU/Charles alumni graph more than through a paid recruiter database.
The DeepMind Edmonton shutdown is a live sourcing event
Here is a detail buried in the TechCrunch write-up that is worth the whole article: the EquiLibre founders were visiting PhD students at DeepMind's first international AI research office in Edmonton, Alberta, where Rich Sutton's group sat. Alphabet shut that office in 2023. That diaspora, researchers with Czech, Canadian, and UK ties who worked on RL at DeepMind Edmonton between roughly 2017 and 2023, is exactly the profile EquiLibre hired first. Some are still at Google, some at other labs, some at trading firms, some in stealth. All of them are warm leads for anyone competing in this space.
Anyone who worked at DeepMind Edmonton between 2017 and 2023 is a warm lead. That office no longer exists.