Member of Technical Staff, Evaluation Execution
Hybrid (Berkeley, CA)$328k - $578.6kEmployeePosted 3mo ago
About the role
Role focused on running evaluation models, communicating results, building evaluation software, and project management to scale METR's frontier AI risk evaluations.
Responsibilities
- Running models on tasks by integrating models into METR's agent scaffolds, running on infrastructure, and verifying results.
- Communicating results and takeaways through useful graphs and written conclusions for diverse audiences (system cards, risk reports, regulators, social platforms).
- Building software to improve evaluations, enabling faster and more informative assessments with platform team support.
- Project management to track multiple concurrent live evaluations and ensure organization amid risk report processes.
- Strong and professional communication with METR leadership, lab contacts, regulators, and other stakeholders.
Qualifications
- Software engineering: strong engineering fundamentals, ability to debug unfamiliar systems, identify and fix performance bottlenecks.
- Speed and scrappiness: ability to move quickly, identify 80/20, and focus on essential work.
- High attention to detail: careful reading, bug spotting in transcripts, attention to fiddly details.
Skills
Required:Software engineeringSystem debugging and performance optimizationCommunication
Benefits
- Retirement: 401(k) with a 100% match up to 8% of salary
- Life Insurance: employer-paid basic life insurance at 1x salary
- Time-off and leave: Unlimited PTO and 21-week parental leave for new parents
- The office: Catered lunch and dinner daily; in-office gym and shower
- Health: Medical, dental and vision coverage with 99% employee premiums paid; Employer funded HRA; Fertility benefits through Carrot
- Additional stipends: Relocation, Wellness, Commuter, Professional development, Work equipment, Management coaching for managers
Last updated Oct 21, 2026. Interested? Apply on the company site.
About METR
Model Evaluation & Threat Research
METR (pronounced 'meter') is a research nonprofit that scientifically measures whether and when AI systems might threaten catastrophic harm to society.
researchevaluationAIthreat
More open roles at METR
- Head of People1w agoBerkeley · Operations
- Member of Policy Staff, Government Affairs1w agoPolicy / Government Affairs
- Chief Information Security Officer1mo agoBerkeley, CA · Engineering & Research
- IT Systems Engineer1mo agoBerkeley, CA · Operations
- Member of Technical Staff, Offensive Security Research1mo agoBerkeley, CA · Engineering & Research