JobEvaluationsInfrastructureMid-level
Evaluations Engineer
Meridian Alignment Lab
Berkeley, CA · Hybrid $150k–$195k 12 Oct 2026
Build the harnesses that let researchers run dangerous-capability evaluations safely and reproducibly.
What you'd do
- Maintain sandboxed eval infrastructure
- Turn threat models into concrete, versioned tests
- Own reproducibility of every published eval result
What they look for
- Strong Python
- Systems or infra experience
- Security-minded
Meridian Alignment Lab
Small empirical lab studying whether oversight methods hold up when the model is smarter than the overseer.
Research lab · 12 people · Philanthropically funded through 2028