Research, Post-Training

Tech Stack

Responsibilities

• Iterate on datasets, training stages, and hyperparameters to develop post-training recipes that determine model behavior and measure performance across evaluations and production.
• Design and build evaluations that accurately capture agent performance, continuously refining them based on observed gaps.
• Deeply understand unexpected training results to identify root causes and apply that knowledge to future problems.
• Apply and advance techniques like RLHF, RLAIF, and constitutional approaches to shape agent reasoning, action, and human collaboration for long-horizon tasks.
• Measure performance scaling with data and compute, and develop new methodologies when existing ones reach their limits.

Culture

Requirements