Researcher: Agent Post-Training, API & Power-Users

📋 Description

• Design and run experiments to improve model behavior in API workflows.
• Build evals, graders, and environments from real developer workflows.
• Turn observed failures into training data and model improvements.
• Partner with API and power-users to identify high-leverage behavior gaps.
• Own end-to-end model behavior projects from data generation to launch readiness.
• Improve large-scale training machinery: speed, reliability, observability.

🎯 Requirements

• Strong technical fundamentals in ML, software engineering, systems, statistics, or applied research.
• Hands-on with LLMs, post-training, evals, graders, synthetic data, or production ML.
• Ability to analyze transcripts/evals and form concrete hypotheses about model behavior.
• Experience tackling ambiguous capability problems where signal is noisy and failures are qualitative.
• Deep care for developer and expert-user experience with API products and agent harnesses.
• Comfortable across research, product, infrastructure, data, evals, and safety; communicate clearly.

🎁 Benefits

• Offers equity
• Hybrid work environment
• Opportunity to work on frontier AI research at OpenAI