Dan Roberts

Also known as Daniel A. Roberts

Dan Roberts is an AI researcher who applies theoretical physics to deep-learning theory and leads Frontiers at OpenAI. He studied physics and electrical and computer engineering at Duke, read for Cambridge’s Part III Mathematical Tripos, and studied quantum information at Oxford as a Marshall Scholar; he earned an MIT PhD in high-energy theoretical physics in 2016. He co-founded Diffeo, whose collaborative-intelligence product became Einstein Relationship Insights after Salesforce acquired the company in 2019, and co-authored The Principles of Deep Learning Theory. He worked at Facebook AI Research and the Institute for Advanced Study, became a Sequoia AI Fellow, and joined the AI Mathematical Olympiad Prize Advisory Committee. In June 2026, he led OpenAI’s Foundations of Reinforcement Learning team; as of October 2026, he leads Frontiers at OpenAI and is a Visiting Scientist at MIT’s Center for Theoretical Physics.

Now at: OpenAI and Massachusetts Institute of Technology
Works: 58
Citations: 5,971
h-index: 19
Badges: Founder
Bridge
Boomerang

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

2009–2019

Hertz, Diffeo and causal learning

He studies physics and electrical and computer engineering at Duke and is a 2009 Hertz Fellow. As a Marshall Scholar, he reads for Cambridge’s Part III Mathematical Tripos and studies quantum information at Oxford. He co-founds Diffeo in 2012, completes a high-energy theoretical-physics PhD at MIT in 2016, and presents “Causality in Physics and Effective Theories of Agency” with Max Kleiman-Weiner at NeurIPS 2018; Salesforce acquires Diffeo in 2019, and its collaborative-intelligence product becomes Einstein Relationship Insights.
2022

A book on deep learning theory

In 2022, Cambridge University Press publishes The Principles of Deep Learning Theory, co-authored by Roberts and Sho Yaida with contributions from Boris Hanin. The book uses theoretical physics to explain how realistic deep neural networks work.
2023

AI Fellow at Sequoia

He joins Sequoia Capital as an AI Fellow in 2023. Earlier in his career, he is a research scientist at Facebook AI Research and a postdoc at the Institute for Advanced Study.
2024

The AIMO Prize committee

In 2024, he joins the advisory committee for the AI Mathematical Olympiad Prize alongside Timothy Gowers, Terence Tao, Po-Shen Loh, and Geoff Smith.
As of October 2026

Leading Frontiers at OpenAI

In June 2026, he leads the Foundations of Reinforcement Learning team at OpenAI. As of October 2026, he leads Frontiers at OpenAI, works on pretraining, reinforcement learning, science and math, and remains a Visiting Scientist at MIT’s Center for Theoretical Physics.

Career

Since 2026OpenAILead, Foundations of Reinforcement Learning team, Lead, Frontiers
Since 2013Massachusetts Institute of TechnologyVisiting Scientist
2017–2018Institute for Advanced StudyPostdoctoral researcher
2012–2019DiffeoCo-founder
FAIR (Meta)Research Scientist