jobs

Agent Post-Training, Frontier Evals and Environments Research

OpenAI's Agent Post-Training team develops the frontier agents used across Codex, ChatGPT, the API, and other products. This role focuses on building evaluation environments and training systems that guide progress toward safe, capable arti

Company

OpenAI

Location

San Francisco, United States

Status

Closed

Posted

2026-09-18T03:33:24.451+00:00

OpenAI's Agent Post-Training team works on the frontier agents behind Codex, ChatGPT and the API. This role builds reinforcement-learning environments and evaluation systems: creating ambitious RL environments, measuring model capabilities and behavior, developing methods to explore model behavior automatically, and improving the reliability and statistical quality of evaluations. It may include steering training runs and building continuous-evaluation systems. Candidates typically need strong foundations in machine learning, software engineering, systems or statistics, plus experience with LLMs, post-training, RLHF/RLAIF, model evaluations, graders, synthetic data, coding or tool-using agents. Hybrid, San Francisco.