OpenAI
Agent Post-Training, Computer Use Research
$380K–$500K
San Francisco · Full-time · Posted Oct 19, 2026
About the job
Train and improve models that reliably operate browsers, desktops, tools, and applications, from research and evaluation through large-scale training and product deployment.
What you'll do
- Design and run experiments and improve the post-training stack, including training data, reward signals, graders, and evaluations.
- Build environments and evaluations to identify failures, then translate findings into training, product fixes, or research directions.
- Collaborate across research, product, infrastructure, and safety teams to improve and launch agent capabilities.
What they ask for
- Strong technical fundamentals in machine learning, software engineering, systems, statistics, or a related field.
- Hands-on experience with one or more of: LLMs, RL, RLHF/RLAIF, post-training, evaluations, graders, synthetic data, model training, coding or tool-using agents, or production ML systems.
Pay
$380K – $500K (USD; base salary or total pay not stated)
Tools
RL, RLHF/RLAIF, LLMs, Codex, ChatGPT
AI focus
Post-train and evaluate agents for computer use, including browser and desktop operation, tool use, complex workflows, and long-horizon tasks; work may include alignment interventions and multi-agent systems.
Model-written summary · the original posting is authoritative