PeopleofSuperintelligence

Researcher

Sylvain Gelly

Sylvain Gelly is a French computer scientist whose research links computer Go, reinforcement learning, and large-scale deep learning at Google. He completed a doctorate at Paris-Sud 11 University in 2007 and co-authored MoGo, the first program to reach dan level in 9×9 Go. With David Silver he published Combining Online and Offline Knowledge in UCT, which received the ICML Test of Time Award in 2017, while he was a researcher on the Google Brain team in Zurich. In 2019 he shared an ICML Outstanding Paper on disentangled representations and co-authored adapter modules for transfer learning, and in 2021 he co-authored the Vision Transformer paper An Image is Worth 16x16 Words. As of October 2026, he is a deep learning researcher at Google Brain in Zurich.

Works
183

Publication numbers from OpenAlex, as of 17 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2007

    A doctorate on computer Go

    A French computer scientist, he defends a doctorate at Paris-Sud 11 University in 2007. The thesis, A Contribution to Reinforcement Learning; Application to Computer Go, comes from the Learning and Optimisation Group in the Laboratoire de recherche en informatique, under Michèle Sebag and Nicolas Bredèche. He co-authors the Go program MoGo, using Monte-Carlo tree search with patterns in the simulations and improvements to UCT.

  2. 2011

    Dan level with MoGo

    With David Silver, he publishes Monte-Carlo tree search and rapid action value estimation in computer Go in Artificial Intelligence in 2011. Two extensions of the search, applied to MoGo, make it the first program to reach dan level in 9×9 Go.

  3. 2017

    An ICML Test of Time Award

    ICML gives the 2017 Test of Time Award to Combining Online and Offline Knowledge in UCT, the 2007 paper with David Silver. In September 2017 he is a researcher on the Google Brain team in the Zurich office.

  4. 2019

    An ICML outstanding paper

    Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations receives an ICML Outstanding Paper award in 2019. He shares the paper with Francesco Locatello, Stefan Bauer, Mario Lucic, Gunnar Rätsch, Bernhard Schölkopf, and Olivier Bachem, and works at Google Research in Zurich.

  5. 2019

    Adapter modules for NLP

    Parameter-Efficient Transfer Learning for NLP appears at the International Conference on Machine Learning in 2019, with him as a co-author. The paper proposes adapter modules that add a few trainable parameters per task. The parameters of the original network remain fixed, and the modules transfer BERT to 26 text classification tasks, including the GLUE benchmark.

  6. 2019

    Google Research Football

    Google Research Football: A Novel Reinforcement Learning Environment appears at AAAI in 2019, with him as a co-author. Agents train to play football in a physics-based 3D simulator that is available under a permissive open-source license, with baseline results for Impala, PPO, and Ape-X DQN.

  7. 2021

    The Vision Transformer paper

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale is presented at ICLR in 2021, with him as a co-author. A pure transformer applied to sequences of image patches, after pre-training on large amounts of data, matches or outperforms convolutional networks on recognition benchmarks while using substantially less computation to train.

  8. As of October 2026

    Google Brain in Zurich

    As of October 2026, he is a deep learning researcher at Google Brain in Zurich.

Sources and credits

Photo: We have no published photo of Sylvain Gelly yet. Send us a free-licensed photo.

  1. Sylvain Gelly - Chess Programming Wiki chessprogramming.org. Checked 16 October 2026.
  2. Monte-Carlo tree search and rapid action value estimation in computer Go research.google. Checked 16 October 2026.
  3. ICML 2017 Awards icml.cc. Checked 16 October 2026.
  4. Seminal Ideas from 2007 research.google. Checked 16 October 2026.
  5. ICML 2019 Awards icml.cc. Checked 16 October 2026.
  6. ETH Zurich scientists among the winning team at world's leading AI conference – Department of Computer Science | ETH Zurich inf.ethz.ch. Checked 16 October 2026.
  7. Parameter-Efficient Transfer Learning for NLP proceedings.mlr.press. Checked 16 October 2026.
  8. Google Research Football: A Novel Reinforcement Learning Environment research.google. Checked 16 October 2026.
  9. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale research.google. Checked 16 October 2026.
  10. Sylvain Gelly - Google Scholar scholar.google.com. Checked 16 October 2026.