# John Schulman

> John Schulman is an American AI researcher, co-founder of OpenAI, and co-founder and chief scientist of Thinking Machines Lab. After studying physics at Caltech and earning a Berkeley PhD in computer science, he developed influential reinforcement-learning methods including TRPO and PPO, led the reinforcement-learning work behind ChatGPT, and co-led OpenAI's post-training team. His work now centers on customizable AI, reinforcement learning, and alignment.

- **Role:** Co-founder and Chief Scientist at Thinking Machines Lab
- **Organizations:** Thinking Machines Lab, Anthropic, OpenAI, University of California, Berkeley, California Institute of Technology
- **Nationality:** American
- **Education:** Bachelor's degree in Physics, California Institute of Technology, PhD in Computer Science, University of California, Berkeley
- **Known for:** Co-founded OpenAI in 2015., Developed Trust Region Policy Optimization and Proximal Policy Optimization., Led the reinforcement-learning team behind ChatGPT and later co-led OpenAI's post-training team.

## Career timeline

- **2005** — Selected for the U.S. Physics Olympiad team.
- **2010** — Graduated from Caltech with a degree in physics and began graduate study at UC Berkeley, initially in neuroscience.
- **2013-2016** — Worked with Pieter Abbeel on robotics and deep reinforcement learning; developed research including TRPO and generalized advantage estimation.
- **2015** — Co-founded OpenAI while finishing his Berkeley doctorate.
- **2016** — Completed his Berkeley PhD, Optimizing Expectations: From Deep Reinforcement Learning to Stochastic Computation Graphs.
- **2017** — Introduced Proximal Policy Optimization, a practical policy-gradient method that became widely used.
- **2018** — Named to MIT Technology Review's Innovators Under 35 list in the Pioneers category.
- **2020-2022** — Shifted into language-model alignment work and contributed to WebGPT, InstructGPT, and the reinforcement-learning work behind ChatGPT.
- **2022-2024** — Co-led OpenAI's post-training team, developing models for ChatGPT and the OpenAI API.
- **2024** — Left OpenAI in August to pursue hands-on alignment research at Anthropic.
- **2025** — Joined the founding team of Thinking Machines Lab as co-founder and chief scientist.
- **2025** — Published LoRA Without Regret with Thinking Machines collaborators and helped launch Tinker, a fine-tuning API.

## Achievements

- Co-founded OpenAI in 2015.
- Developed Trust Region Policy Optimization and Proximal Policy Optimization.
- Led the reinforcement-learning team behind ChatGPT and later co-led OpenAI's post-training team.
- Co-authored influential work on generalized advantage estimation, WebGPT, InstructGPT, and reinforcement learning from human feedback.
- Named an MIT Technology Review Innovator Under 35 in 2018.
- Co-founded Thinking Machines Lab and serves as its chief scientist.

## Latest updates

- **2026-07** — Thinking Machines Lab released Inkling, an open-weights model designed for customization, and published its case for widening model access in stages.
- **2026-05** — Thinking Machines Lab previewed interaction models for continuous, real-time human-AI collaboration.
- **2026-03** — Thinking Machines Lab announced a long-term strategic partnership with NVIDIA for gigawatt-scale computing infrastructure.
- **2025-10** — Thinking Machines Lab announced Tinker, an API for fine-tuning language models.
- **2025-09** — Schulman published LoRA Without Regret with collaborators at Thinking Machines Lab.

## Links

- Website: https://joschu.net/
- Twitter/X: https://x.com/johnschulman2
- GitHub: https://github.com/joschu

---

Profile page: https://yespress.io/john-schulman
Published by YesPress — https://yespress.io
Last updated: 2026-09-26