jobs

Agent Post-Training, API & Power Users

The Agent Post-Training team creates the models behind OpenAI's agents in Codex, ChatGPT, the API, and other frontier products. The team builds the data, environments, graders, training methods, and feedback loops that shape how agents code

Company

OpenAI

Location

San Francisco, United States

Status

Closed

Posted

2026-09-13T14:51:18.676+00:00

OpenAI's Agent Post-Training group builds the models behind its agents in Codex, ChatGPT and the API. On the API and power-users team, this role improves agentic models for API developers and expert users: designing evaluations from real developer workflows, building training environments and graders, running experiments on function calling, tool use, long-horizon execution, factuality and error recovery, and turning failures into training data or post-training interventions. Candidates need strong fundamentals in machine learning, software engineering, systems, statistics or applied research, plus experience with LLMs, post-training, RL/RLHF, evaluations or tool-using agents.