Oriol Vinyals
Co-founder, Discovery Loop
Co-invented sequence-to-sequence learning, led AlphaStar and co-led Gemini before co-founding Discovery Loop in 2026.
Oriol Vinyals grew up in Sabadell, near Barcelona, and took a double degree in mathematics and telecommunications engineering at the Universitat Politècnica de Catalunya before graduate study in California, finishing with a PhD at the University of California, Berkeley. At Google Brain he wrote, with Ilya Sutskever and Quoc Le, the September 2014 paper "Sequence to Sequence Learning with Neural Networks," which used one neural network to read a sentence and another to write its translation. On English-to-French translation it scored 34.8 BLEU, above the 33.3 of a phrase-based system on the same data. In March 2015 he, Geoffrey Hinton and Jeff Dean described knowledge distillation, training a small network to imitate a large one.
At DeepMind in London he led AlphaStar, a StarCraft II agent trained with imitation and multi-agent reinforcement learning. In January 2019 it beat two professional players, and in October 2019 Nature reported that it had reached Grandmaster level on Battle.net, the game's public online ladder, ranking above 99.8 percent of active players while limited to a camera view and action rate comparable to a human's.
He became a vice president of research, and the December 2023 Gemini technical report named him and Dean the overall technical leads of Gemini, the model family built after DeepMind merged with Google Brain. In November 2025 his alma mater made him an honorary doctor.
On 5 August 2026, the day Demis Hassabis stepped back as chief executive, Vinyals left Google DeepMind to co-found Discovery Loop with Dean, Sanjay Ghemawat and Quoc Le, a public benefit corporation that aims to automate the experimental loop of research. Speaking at the Agentic AI Summit in September 2026, he said AI systems would increasingly improve themselves but that the hardest step to automate was "research taste," the judgment of which ideas are worth pursuing.
Known for
Sequence-to-sequence learning
The 2014 encoder-decoder method, with Sutskever and Le, that translated English to French with a single neural network reading and writing whole sentences.
Knowledge distillation
A 2015 paper with Hinton and Dean on training small models to reproduce the outputs of large ones.
AlphaStar
DeepMind's StarCraft II agent, which reached Grandmaster level in 2019 in a real-time game with hidden information.
Career
-
2013-2016
Google Brain
Research scientist
-
2016-2026
Research scientist, then VP of Research and Gemini co-lead
-
2026-present
Discovery Loop
Co-founder
Sources
- Sequence to sequence learning with neural networks (arXiv, September 2014)
- Distilling the knowledge in a neural network (arXiv, March 2015)
- Grandmaster level in StarCraft II using multi-agent reinforcement learning (Nature, October 2019)
- AlphaStar, Grandmaster level in StarCraft II (Google DeepMind, October 2019)
- Gemini, a family of highly capable multimodal models (Gemini Team, arXiv, December 2023)
- Oriol Vinyals investit doctor honoris causa per la UPC (CFIS, UPC, December 2025)
- Jeff Dean and other top AI researchers are leaving Google to launch their own startup (TechCrunch, August 2026)
- Ex-DeepMind VP Vinyals says AI self-improvement is coming but won't trigger an intelligence explosion (The Decoder, September 2026)
Brief profiles cover a person's professional record: roles, work and public statements, each from a source listed here. Full profiles, with positions across the debates, are kept for the people the site follows most closely.