Researcher, Alignment Cot Monitorability

Researcher on the Alignment team at OpenAI studying chain-of-thought (CoT) monitorability in frontier reasoning models. The role involves designing and running experiments to understand model monitorability, investigating training interventions, building evaluations, and translating findings into practical recommendations.

Skills and focus areas

• machine learning
• model behavior
• alignment
• interpretability
• empirical ml
• chain-of-thought interpretability
• model training
• investigative model-behavior
• oversight
• reward functions
• environments
• monitoring