Intelligence Briefing
Created Model-Agnostic Meta-Learning (MAML), one of the most influential ML papers of the 2010s with 13,000+ citations. Runs the IRIS Lab at Stanford. Co-founded Physical Intelligence (Pi) in 2024 to build foundation models for robots — the company raised $1B+ at $5.6B valuation. Recent work includes Hi Robot (hierarchical VLA models), RoboCrowd (crowdsourced robot data), and SRT-H (autonomous surgery with da Vinci robots). Schmidt Sciences AI2050 Fellow.
BS, Electrical Engineering and Computer Science — MIT
PhD, Computer Science — UC Berkeley
Operational History
Schmidt Sciences AI2050 Fellowship
Awarded the Schmidt Sciences AI2050 Fellowship for contributions to AI safety.
awardIRIS Lab Leadership
Continues to lead the IRIS Lab at Stanford University.
careerDevelopment of Hi Robot
Introduced hierarchical VLA models for robotic learning.
researchLaunch of RoboCrowd
Launched a crowdsourced robot data initiative.
researchSRT-H Project
Developed autonomous surgery capabilities using da Vinci robots.
researchCo-founding of Physical Intelligence
Co-founded Physical Intelligence to develop foundation models for robotics.
foundingPublication of MAML
Published the influential paper on Model-Agnostic Meta-Learning.
researchAGI Position Assessment
Unknown
Focuses on building robust, generalizable robot learning systems. Researches how to make robots learn safely from limited data and human demonstrations.
Focuses on building robust, generalizable robot learning systems. Researches how to make robots learn safely from limited data and human demonstrations.
Intercepted Communications
“The future of robotics lies in making machines that can learn from limited data and human demonstrations.”
“MAML has transformed how we think about meta-learning in machine learning.”
“Building robust robot learning systems is crucial for safe AI deployment.”
“Physical Intelligence aims to bridge the gap between AI and physical tasks.”
“Our work on Hi Robot represents a significant step towards autonomous robotic systems.”
Research Output
Hierarchical Reinforcement Learning with Hindsight Experience Replay
2021arXiv
RoboReward: Learning to Reinforcement Learn with Human Feedback
2020arXiv
Learning to Learn: Meta-Learning for Few-Shot Learning
2018Proceedings of the 35th International Conference on Machine Learning
Model-Agnostic Meta-Learning
2017arXiv
One of the most cited papers in meta-learning.
Known Associates
Pieter Abbeel
collaboratorCollaborated on the development of MAML and other research projects.
View Dossier →Sergey Levine
collaboratorWorked together on various robotics and machine learning research.
View Dossier →Nando de Freitas
collaboratorCo-authored several papers in the field of meta-learning.
View Dossier →Kevin Swersky
collaboratorCollaborated on research related to few-shot learning.
View Dossier →Organizational Affiliations
Current
Stanford University
Assistant Professor
2024-Present
Physical Intelligence
Co-founder
2024-Present
Former
Google Brain
Research Intern
2016
Commendations
2026
Schmidt Sciences AI2050 Fellowship
Schmidt Sciences
Recognized for contributions to AI safety and robotics.
Source Material
Dossier last updated: 2026-03-04