Research Engineer, Post-Training

About this role

Harvey seeks a Research Engineer to lead post-training efforts that improve AI agent performance on legal work. You'll design training experiments, build evaluation systems, and collaborate with researchers to optimize models through feedback loops and domain-specific optimizations.

What you'll do

• Drive post-training experiments balancing performance, cost, latency, and security trade-offs
• Optimize agent systems including skills, tools, retrieval strategies, and validation loops for legal tasks
• Design reliable grading and reward systems for evaluation and training iteration
• Analyze agent behavior patterns and convert insights into training data or harness improvements
• Collaborate with Harvey researchers and external partners on experiment design and model improvements

What they're looking for

• Post-training expertise (SFT, preference optimization, RLHF, reward modeling, distillation)
• Strong Python and research-engineering ability
• Model behavior analysis and failure mode identification
• Self-management of ambiguous applied research projects
• Data and evaluation infrastructure building
• Distributed training and GPU workload experience
• Clear communication across research, engineering, and product teams

Benefits

• Competitive salary: $231,000 - $340,000
• Work on frontier AI transforming legal and professional services
• Opportunity for significant personal and professional growth
• Collaborate with world-class researchers and engineers
• Scale impact with 1500+ customers in 60+ countries