Research Engineer / Scientist, Reinforcement Learning

About the job

Responsibilities

• Own and advance a research agenda to enhance model capabilities in long-horizon legal reasoning and agentic workflows.
• Design and execute post-training experiments using SFT and RL, building upon leading open-weight models.
• Develop reward functions and verifiers for complex, semi-verifiable tasks, including creating rubrics, learned preference models, and expert preference data, and analyzing their limitations.
• Manage and troubleshoot large-scale training runs, addressing issues like divergence, reward hacking, and instability.
• Create and utilize rigorous evaluation methods to measure improvements and determine which models to deploy.
• Design, implement, test, and debug code across the research technology stack.

Requirements

• Hands-on experience with post-training LLMs using SFT and RL, including preference optimization, with direct responsibility for training runs and outcomes.
• Deep understanding of ML fundamentals and proven experience in reliably executing distributed training runs.
• Proficiency in Python engineering, capable of writing production-quality code and managing end-to-end systems.
• Research experience in RL for language models or reward modeling, with the ability to rapidly translate experiments into deployed improvements.
• Publications are welcome but not required.