Software Engineer RL Data

You will design and iterate task sets using traces and evaluations to improve coding-agent capabilities. You will investigate agent failures, build systems that surface related behavior, improve rewards and environments, and partner with research to assess dataset quality.

Responsibilities

• Design and iterate task sets that teach agent capabilities
• Analyze agent traces to identify failure modes and build systems that surface similar behavior
• Turn one-off recipes into reusable rewards, environments, and data-quality improvements
• Partner with research to assess whether datasets teach intended capabilities

Requirements

• Strong software engineering fundamentals
• Infrastructure, data, or distributed systems background
• Ability to turn messy real-world agent behavior into a dataset or tool