Jerry Tworek
Scholar

Jerry Tworek

Google Scholar ID: ZPuESCQAAAAJ
OpenAI
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
34,235
 
H-index
11
 
i10-index
11
 
Publications
14
 
Co-authors
0
 
Contact
TwitterOpen ↗GitHubOpen ↗
Publications
2 items
GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks
2025
Cited
0
Competitive Programming with Large Reasoning Models
2025
Cited
0
Resume
Research Experience
  • Led multiple core research initiatives at OpenAI, including o1, o3, GPT-4, Codex, GitHub Copilot, ChatGPT, and its plugin systems.
  • Led the team that pioneered reinforcement learning for reasoning in LLMs, resulting in the OpenAI o1 model (September 2024).
  • Directed research that enabled GPT-4 to become the world’s strongest model for programming challenges (March 2023).
  • Pioneered dialog-based code-writing agents that made ChatGPT widely useful for programmers globally (November 2022).
  • Initiated and led code-focused LLM research at OpenAI since 2021, serving as a primary contributor to Codex and GitHub Copilot.
  • Contributed a core breakthrough in the 2019 Rubik’s Cube robotic hand project using sim2real transfer learning.