Michal Valko
Also known as 马悯伽
Michal Valko is an AI researcher whose work runs from bandit algorithms to Llama 3. He earned a PhD at the University of Pittsburgh in 2011 and became a tenured researcher at Inria in 2012. With Rémi Munos, he helped found the Google DeepMind lab in Paris, and he co-authored the self-supervised BYOL method in 2020. A paper he co-wrote on zero-sum imperfect-information games won an ICML Outstanding Paper Award in 2023. He became a principal Llama scientist at Meta in 2024, and he is a founding researcher at Isara Labs, which he joined in 2025.
- Now at
- Isara Labs
- Works
- 195
- Citations
- 5,585
- h-index
- 25
Publication numbers from OpenAlex, as of 15 October 2026.
Timeline
-
2005–2011
Degrees in Bratislava and Pittsburgh
He earns a master's degree in computer science at Comenius University in Bratislava in 2005. He completes a PhD at the University of Pittsburgh in 2011.
-
2012–2021
A tenured post at Inria
He receives a tenured research post at Inria in 2012. Inria gives him its scientific excellence award for 2014–2017 and again for 2018–2021.
-
2014
Implicit exploration for bandits
With Tomáš Kocák, Gergely Neu and Rémi Munos, he publishes a paper on implicit exploration for bandit problems with side observations in the NIPS 2014 proceedings.
-
2019–2024
Co-founding DeepMind Paris
Alongside Rémi Munos, he helps found the Google DeepMind lab in Paris. He works there until 2024.
-
2020
Bootstrap Your Own Latent
He co-authors Bootstrap Your Own Latent (BYOL), a self-supervised representation-learning method presented at NeurIPS in 2020.
-
2023
An ICML Outstanding Paper Award
Together with his co-authors, he receives an ICML Outstanding Paper Award in 2023 for Adapting to game trees in zero-sum imperfect information games.
-
2024
Principal Llama scientist at Meta
In 2024, he becomes a principal Llama scientist at Meta, building an online reinforcement learning stack for Llama 3. He co-authors Nash Learning from Human Feedback, an alternative to standard reinforcement learning from human feedback.
-
Since 2025
Founding researcher at Isara Labs
In 2025, he joins Isara Labs as a founding researcher, where he works on a new generation of large language models.
Sources and credits
Photo: We have no published photo of Michal Valko yet. Send us a free-licensed photo.
- Michal Valko - Wikipedia
- ETD | Adaptive Graph-Based Algorithms for Conditional Anomaly Detection and Semi-Supervised Learning | ID: b1d8ccc1-8a99-4765-9841-1817d783cc34 | Hyku Commons
- Michal Valko 马悯伽 | ELLIS Fellow | AI Researcher, RL and LLMs
- Michal VALKO - Université de Lille
- Efficient learning by implicit exploration in bandit problems with side observations
- Bootstrap Your Own Latent - A New Approach to Self-Supervised Learning
- ICML 2023 Awards
- Speakers - Michal Valko
- Nash Learning from Human Feedback