PeopleofSuperintelligence

Researcher

Theophane Weber

Theophane Weber is a researcher whose work spans probabilistic inference, visual reasoning, reinforcement learning and mathematical reasoning. He earned an MS and PhD at MIT's Operations Research Center, then worked at Lyric Labs before joining DeepMind. He coauthored research on automated inference, generative scene understanding, physical reasoning and agents that plan with learned models. In 2024 and 2025, he contributed to Project Big Sleep's vulnerability research and to Gemini Deep Think's gold-medal-level result at the International Mathematical Olympiad. As of October 2026, he is a Research Scientist and Team Lead in Gemini at Google DeepMind, focused on reinforcement-learning science, language-model reasoning and AI-based scientific discovery.

Badges

Timeline

  1. 2010

    A thesis on graphical models

    In 2010, he completes a PhD thesis at MIT's Operations Research Center on correlation decay and decentralized optimization in graphical models. He earns an MS and PhD there, with David Gamarnik and John Tsitsiklis co-advising his PhD and Jérémie Gallien advising his MS.

  2. 2013

    Probabilistic program inference

    In January 2013, he coauthors Automated Variational Inference in Probabilistic Programming with David Wingate. Their algorithm uses stochastic gradients to automate approximate inference in probabilistic programs.

  3. 2016

    Attend, Infer, Repeat

    Before DeepMind, he works at Lyric Labs, which Analog Devices acquires. In December 2016, he coauthors Attend, Infer, Repeat, a generative model that learns to count, locate and classify scene objects without supervision.

  4. June 2017

    Visual Interaction Network

    In June 2017, with Nicholas Watters and Daniel Zoran, he helps develop the Visual Interaction Network. It infers physical-object states from a few video frames and predicts their positions hundreds of steps into the future.

  5. July 2017

    Imagination-Augmented Agents

    In July 2017, he coauthors Imagination-Augmented Agents for Deep Reinforcement Learning with colleagues. The agents interpret predictions from a learned environment model as implicit plans, while an imagination encoder extracts information useful for future decisions.

  6. 2024–2025

    Project Big Sleep

    In November 2024, Google Project Zero and Google DeepMind describe Big Sleep as a collaboration whose AI agent found an exploitable SQLite stack buffer underflow in early October 2024. Weber is listed among Big Sleep contributors by November 2024. In July 2025, the team pre-empts an in-the-wild zero-day exploit.

  7. 2025

    Gemini Deep Think at IMO

    In July 2025, Gemini Deep Think solves five of six IMO problems, scores 35 of 42 points and reaches gold-medal level. At DeepMind, he is a key contributor to the IMO and DeepThink efforts and co-invents the training technique behind the Gemini model.

  8. As of October 2026

    Team lead in Gemini

    As of October 2026, he is a Research Scientist and Team Lead in Gemini at Google DeepMind. His focus spans reinforcement-learning science, language-model reasoning and AI-based scientific discovery.

Career

  1. NowGoogle DeepMindResearch Scientist and Team Lead in Gemini

Education

  1. 2010Massachusetts Institute of TechnologyMS

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Theophane Weber. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

Sources and credits

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

  1. PhD and Masters Theses – Operations Research Center orc.mit.edu. Checked 21 October 2026.
  2. ORC Seminar: Scaling AI Capabilities in Mathematical Reasoning calendar.mit.edu. Checked 21 October 2026.
  3. [1301.1299] Automated Variational Inference in Probabilistic Programming arxiv.org. Checked 21 October 2026.
  4. DeepMind Papers @ NIPS (Part 3) deepmind.google. Checked 21 October 2026.
  5. A neural approach to relational reasoning deepmind.google. Checked 21 October 2026.
  6. [1707.06203] Imagination-Augmented Agents for Deep Reinforcement Learning arxiv.org. Checked 21 October 2026.
  7. Agents that imagine and plan deepmind.google. Checked 21 October 2026.
  8. From Naptime to Big Sleep: Using Large Language Models To Catch Vulnerabilities In Real-World Code projectzero.google. Checked 21 October 2026.
  9. Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad deepmind.google. Checked 21 October 2026.