Software Engineer

You'll design, build, and quality-assure RL tasks. Each task is a self-contained software engineering challenge with a prompt, an environment, and an automated grader. You own the full lifecycle: ideation, grading infrastructure, running frontier models against the task, failure analysis, and iteration. At this level, we expect you to consistently produce tasks that target meaningful capability gaps in frontier models, and to develop a strong sense for what makes a task informative versus merely difficult.

You will use coding agents heavily, and a large part of the job is directing them well, evaluating their output, and knowing when they are failing in subtle ways. You may also contribute to shared infrastructure: improving our build pipeline, automating parts of QA, or building tooling for other engineers.

What makes someone good at this

Strong technical fundamentals combined with a well-calibrated intuition for AI model behavior. You need to anticipate where a model will take shortcuts, distinguish genuine capability gaps from grader issues, and understand how a model will interpret a prompt. At this level, we expect extensive familiarity with what frontier coding agents can and can't do.