Evaluation Engineer

The role Research designs evaluations; you make them run reliably. You’ll build the harnesses, graders and reporting that let a model be evaluated on hundreds of tasks across our environments, with results anyone can reproduce. What you’ll own The evaluation harness: task loading, environment orchestration, model adapters, scoring.Graders and verifiers implemented with the environment and verifier engineers.Runs that are deterministic, resumable and cheap to repeat.Dashboards and reports that researchers and customers read.Discrepancies between runs, chased until they’re explained. What you bring Strong software engineering in Python with an eye for reliability.Model APIs, batch systems and asynchronous processing.You care about reproducibility and know how to achieve it.Comfort reading research code and making it production-grade. Helpful, not required Open evaluation frameworks.Containers and cloud infrastructure.