5-7 years of experience in post-training LLMs with SFT and RL
Strong understanding of machine learning fundamentals
Proficient in Python with a focus on production-quality code
Experience in distributed training runs
Research background in RL for language models or reward modeling
Responsibilities
Own and drive a research agenda for enhancing legal reasoning models
Design and execute post-training experiments using SFT and RL
Develop reward functions and verification methods for semi-verifiable tasks
Debug and manage large-scale training processes
Create rigorous evaluations to track model improvements
Implement and test code across the research stack
Benefits
Collaborative environment with a high-bar team
Opportunity to influence research direction
Hands-on experience with cutting-edge technology
Potential for rapid deployment of research outcomes
Access to a wealth of resources for professional growth
Full Job Description
Role Overview
As a Member of Technical Staff, Modeling in Harvey Research, you'll own how we adapt and improve leading open-weight models for legal work.
We recently post-trained our first model built for legal work, and we're now building the team to scale up our post-training efforts. You'll own the post-training recipe across SFT and RL, including reward design and verification for challenging, semi-verifiable legal tasks. You'll help set the research direction and carry ideas from experiments through training runs to models used by customers. You'll join a small, high-bar team of hands-on engineers and researchers, where each person owns a large part of the work.
We're looking for research engineers and scientists who have gotten training runs working at scale, are immersed in RL for language models, and want their work to reach users quickly.
Learn more about our work at harvey.ai/research.
What You'll Do
Own and pursue a research agenda to improve model capability on long-horizon legal reasoning and agentic workflows.
Design and run post-training experiments across SFT and RL, building on leading open-weight models.
Develop reward functions and verifiers for semi-verifiable tasks, including rubrics, learned preference models, and expert preference data, and study where they break.
Run and debug training at scale, addressing divergence, reward hacking, and instability.
Build and use rigorous evals to track improvements and decide what ships.
Design, implement, test, and debug code across our research stack.
What You Have
Hands-on experience post-training LLMs with SFT and RL, including preference optimization, and direct ownership of training runs and results.
Deep ML fundamentals and experience getting distributed training runs working reliably.
Strong engineering skills in Python, with production-quality code and end-to-end system ownership.
Research experience in RL for language models or reward modeling, with the ability to move quickly from experiments to shipped improvements. Publications welcome, not required.
We are hiring for this role across multiple levels.
Compensation
$230,000 - $340,000 USD
Depending on your location, an Applicant Privacy Notice may apply to you. You can find all of our Applicant Privacy Notices here.
