Researcher, Agent Post-Training, Personality

Researcher for the Agent Post-Training Personality team at OpenAI to help make OpenAI’s agents exceptional collaborators. Study what makes an agent thoughtful, clear, perceptive, appropriately proactive, and easy to work with, then translate those insights into evals, training data, reward signals, and model improvements. Responsibilities include developing rigorous understandings of collaborative behaviors across professional, creative, technical, and everyday contexts; turning qualitative judgments about model behavior into falsifiable hypotheses, evals, graders, and training interventions; studying user signals to determine which behaviors build trust and satisfaction; producing high-quality rollout data and preference data; improving reward models and RL objectives; collaborating with pretraining and post-training teams to shape upstream data choices; building pipelines for updating training data; partnering with product teams to turn consumer insight into model improvements and validating them in real workflows; owning projects end to end from observation to launch.