Researcher, Context - Agent Post-Training

Run experiments and improve post-training systems to scale context for agent models.

• •Work on Agent Post-Training to design experiments and improve scaling of compute on context, turning failures into training data and shipping improvements into frontier models.
• •Key Responsibilities Design and run experiments to improve context compute scaling.
• •Own post-training stack improvements: data pipelines, RL, graders, reward signals, and evals.
• •Build evals/environments and convert failures into training data or product fixes.
• •Collaborate with product, infrastructure, and safety teams to decide model run inclusions.
• •Requirements Strong technical fundamentals in ML, systems, statistics, or software engineering.
• •Hands-on experience with LLMs, RL, RLHF/RLAIF, post-training, or production ML systems.
• •Ability to run end-to-end experiments: hypothesis, pipeline, run, analyze, and act.
• •Cross-functional communication across research, product, infra, data, and safety.