OpenAI
Mid
Research scientist shipping post-training improvements for OpenAI's agentic models
Interview for OpenAI's Agent Post-Training team, which builds the RL, data, evals, and reward-signal pipelines behind agents in Codex, ChatGPT, and the API. Expect open-ended research-engineering questions on turning vague agent-behavior failures into concrete experiments across coding, tool use, computer use, and multi-agent coordination.
Step into this interview
Free · a live voice mock calibrated to this exact role
What this interview tests
- RL / post-training for LLMs
- Eval and environment design
- Agentic behaviors: tool use, computer use, multi-agent coordination
- Research-to-production experiment velocity
- Cross-functional translation of product signal into model training
- Calibrated reasoning and factuality
Common question themes
Turning a vague agent failure mode into a concrete experiment and training fix
Designing evals/graders that reveal model weaknesses before shipping
RL/RLHF/RLAIF experience and reward-signal design tradeoffs
What makes an agent 'reliable and tasteful' — defend with a real example
Owning a post-training improvement from idea through integration and launch
Improving reliability, reproducibility, or cost of large training runs
Related interviews
OpenAI
Senior
Forward Deployed Engineer
OpenAI
Senior
Data Scientist, Identity
OpenAI
Senior
AI Deployment Engineer
Cohere
Senior
Senior Software Engineer, Agent Infrastructure
Figma
Senior
Software Engineer, AI Platforms
Mid