Pushmeet Kohli
Vice President of Research, Google DeepMind, and Chief Scientist, Google Cloud
Leads Google DeepMind's AI for Science unit, whose work spans AlphaFold, AlphaTensor, SynthID and AlphaEvolve.
Pushmeet Kohli grew up in Dehradun, in the Himalayan foothills of northern India, and moved to Britain to study, taking a PhD in computer vision at Oxford Brookes University under Philip Torr; his thesis on graph-cut methods won the Sullivan Doctoral Thesis Award. He then spent nearly 11 years at Microsoft Research, ending as director of research in its Cognition group, which worked on systems able to carry out most of the tasks people do.
He joined DeepMind in 2017 and soon set up its Safe and Reliable AI team, which studied the problems that arise when machine learning models are deployed, as distinct from the long-run AGI safety work Shane Legg had led "from day zero." He also built the lab's science program. As head of what became the AI for Science and Strategic Initiatives unit he oversaw the team around AlphaFold, and is a co-author of the 2021 AlphaFold paper. His unit went on to AlphaTensor (October 2022), which found faster algorithms for multiplying matrices; FunSearch (December 2023), which used a language model to find new constructions in combinatorics; and, in 2025, AlphaEvolve, a Gemini-based agent that designs algorithms. He co-led SynthID, the watermark Google launched for AI images in August 2023 and extended to text in a Nature paper in October 2024.
TIME named him to its 2023 list of the 100 most influential people in AI, describing him as, in his own words, "cautiously optimistic or optimistically cautious." Kohli is a vice president of research and Distinguished Scientist at Google DeepMind, and by 2026 his Google biography also listed him as Chief Scientist of Google Cloud. His June 2026 talk at SXSW London was billed under his Google DeepMind title, and no departure had been reported by September 2026.
Known for
AI for Science
Built the Google DeepMind unit that applies AI to open scientific problems, including protein structure (AlphaFold), matrix multiplication (AlphaTensor) and mathematical discovery (FunSearch, AlphaEvolve).
SynthID
Co-led the watermarking system that marks AI-generated images and text so they can be detected later; the text version was described in Nature in October 2024 and released as open source.
Career
-
2006-2017
Microsoft Research
Researcher, then Director of Research, Cognition group
-
2017-present
Head of Safe and Reliable AI, then Vice President of Research and head of AI for Science and Strategic Initiatives
Sources
- Pushmeet Kohli (Google Research profile)
- Pushmeet Kohli, author page (Google blog)
- Pushmeet Kohli, TIME100 AI (TIME, September 2023)
- Identifying AI-generated images with SynthID (Google DeepMind, August 2023)
- Scalable watermarking for identifying large language model outputs (Nature, October 2024)
- Discovering faster matrix multiplication algorithms with reinforcement learning (Nature, October 2022)
- Mathematical discoveries from program search with large language models (Nature, December 2023)
- Pushmeet Kohli, speaker (SXSW London, June 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.