Jerry Tworek is Founder of Core Automation, an AI startup pioneering continual learning models that require 100 times less training data than current AI systems. He previously served as VP of Research at OpenAI for nearly 7 years, where he led breakthrough developments in reasoning models.

Jerry earned his Master's degree in Applied Mathematics from University of Warsaw and spent 5 years as a quantitative trader in Amsterdam analyzing futures markets using optimization theory. He joined OpenAI in 2019 as a research scientist.

Jerry led the development of OpenAI's o1 and o3 reasoning models, achieving doctorate-level performance in physics, chemistry, and biology. He was the primary researcher behind Codex, which became the foundation for GitHub Copilot, and contributed to GPT-4 and ChatGPT.

Core Automation is seeking between 500 million and 1 billion dollars in funding to develop Ceres, a single model that learns continually in production and challenges the dominant pretrain-fine-tune pipeline. The company launched in January 2026 with ambitions ranging from automating factories to eventually terraforming planets.

At OpenAI, Jerry's work on the first coding models started the LLM coding revolution and his team established the new paradigm of scaling training and inference compute for reasoning models. He also contributed to solving Rubik's Cube with a robotic hand using reinforcement learning and sim2real transfer.