Software Engineer, RL Data

About this role

Cursor is hiring a Software Engineer for their RL Data team to design training tasks, rewards, and environments that improve their coding agents. You'll analyze agent behavior, create datasets, and build reusable systems that make models more effective at coding tasks.

What you'll do

• Design task sets that teach specific agent capabilities and iterate based on model performance traces
• Analyze agent traces to identify failure modes and build systems to surface similar patterns
• Convert one-off reward/environment recipes into reusable infrastructure for other teams
• Collaborate with research to validate that datasets teach intended capabilities
• Manage data quality and create evaluation frameworks for agent training
• Build tooling to process and analyze real-world agent behavior at scale

What they're looking for

• Software engineering fundamentals and writing performant, reliable code
• Data pipeline and infrastructure design
• Distributed systems experience
• Reinforcement learning (preferred but not required)
• Python or similar production languages
• Experience with dataset curation and evaluation metrics
• Problem decomposition and task specification
• Analytical thinking for identifying patterns in complex data