PeopleofSuperintelligence

Researcher

Tianhe Yu

Also known as Tianhe (Kevin) Yu

Tianhe Yu is a computer scientist whose research connects robot learning, reinforcement learning and multimodal models. At UC Berkeley, he earned honorable mention for the CRA Undergraduate Researcher Award in 2017 and a bachelor's degree with highest honors; he completed a Stanford computer science PhD with Chelsea Finn in 2022. His work includes one-shot robot imitation, the Meta-World benchmark, PCGrad, PaLM-E and the Gemini model family. At Google DeepMind, he contributed to post-training for the 2025 IMO version of Gemini Deep Think and foundational research behind its ICPC version; both reached gold-medal-level performance. Since 2025, he works at Meta after leaving Google DeepMind as the company builds Superintelligence Labs.

Now at
Meta

Timeline

  1. 2017–2022

    Berkeley honors and a PhD

    As a UC Berkeley senior in computer science, statistics and math, he earns honorable mention for the CRA Undergraduate Researcher Award in 2017. He earns a bachelor's degree with highest honors in computer science, applied mathematics and statistics at Berkeley, and completes a Stanford computer science PhD in 2022 with Chelsea Finn as his adviser. His research interests span machine learning, perception, control, offline reinforcement learning, multi-task learning and meta-learning.

  2. 2017

    One-shot robot imitation

    At CoRL 2017, he coauthors One-Shot Visual Imitation Learning via Meta-Learning, a method that lets a robot acquire a new skill from one visual demonstration.

  3. 2019

    A 50-task robot benchmark

    He coauthors Meta-World in 2019, an open-source simulated benchmark for meta-reinforcement learning and multi-task learning with 50 robotic manipulation tasks. It evaluates how algorithms generalize to held-out tasks.

  4. 2020

    PCGrad for multi-task learning

    At NeurIPS 2020, he coauthors Gradient Surgery for Multi-Task Learning and introduces PCGrad. The method projects task gradients to reduce interference between conflicting objectives.

  5. 2023

    PaLM-E and the Gemini report

    In 2023, he coauthors PaLM-E, an embodied multimodal language model that incorporates real-world sensor input into language models for embodied tasks. He also coauthors the Gemini report on models that understand images, audio, video and text.

  6. 2025

    Gemini Deep Think at the IMO

    In 2025, DeepMind names him among the corresponding post-training contributors for its advanced Gemini Deep Think model at the IMO. The model solves five of six problems perfectly, earns 35 points and reaches gold-medal-level performance.

  7. 2025

    Gemini Deep Think at the ICPC

    For the 2025 ICPC, DeepMind names him among researchers behind foundational work in Gemini post-training, Thinking and Coding. An advanced Gemini 2.5 Deep Think solves 10 of 12 problems and achieves gold-medal-level performance.

  8. Since 2025

    Joining Meta

    In July 2025, he joins Meta from Google DeepMind as the company builds Superintelligence Labs.

Career

  1. Since 2025Meta
  2. Until 2025Google DeepMind

Education

  1. University of California, BerkeleyBachelor's degree
  2. Stanford UniversityPhD

From the biography.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Tianhe Yu. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  2. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  3. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  4. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Meta. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Four EECS undergraduate researchers recognized by CRA eecs.berkeley.edu. Checked 21 October 2026.
  2. Tianhe (Kevin) Yu cs.stanford.edu. Checked 21 October 2026.
  3. An in-depth analysis of the 100-person map of embodied intelligence in one article: The battle for AI talent among technology giants coinlive.com. Checked 21 October 2026.
  4. Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning arxiv.org. Checked 21 October 2026.
  5. Gradient Surgery for Multi-Task Learning arxiv.org. Checked 21 October 2026.
  6. PaLM-E: An Embodied Multimodal Language Model arxiv.org. Checked 21 October 2026.
  7. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 21 October 2026.
  8. Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad deepmind.google. Checked 21 October 2026.
  9. Gemini achieves gold-medal level at the International Collegiate Programming Contest World Finals deepmind.google. Checked 21 October 2026.
  10. Meta ficha a tres investigadores de Google que llevaron la IA al podio parentesis.media. Checked 21 October 2026.
  11. Meta, Microsoft hire top AI talent from DeepMind economictimes.indiatimes.com. Checked 21 October 2026.