Research Scientist or Director

About the role

AVERI is a US-based nonprofit focused on third party auditing of frontier AI for safety and security. This role owns research on evaluation and auditing of frontier AI systems, including design of experiments, reporting, and contributing to governance standards.

What you'll do

• Own a line of research on frontier AI evaluation from scoping to delivery
• Design and run evaluations of closed and open-weight AI systems, including pilot audits
• Develop audit methods and evaluation protocols for model behavior and safety properties
• Build reproducible experimental pipelines to test models and interpret results
• Produce rigorous audit reports and translate findings into actionable recommendations
• Publish and present research through papers, white papers, open source tools, talks, and writing

What they're looking for

• Independent AI/ML research experience with a focus on safety, robustness, or security (roughly three+ years post PhD or equivalent)
• Strong technical judgment to design rigorous empirical studies and reason about measurement and uncertainty
• Experience with large language models or modern AI systems in safety contexts
• Ability to plan research projects from open questions to tractable plans and work well remotely
• Motivation for AI safety, governance, and third-party auditing
• Nice to have: prior work in AI auditing, interpretability, privacy, cybersecurity, or reproducible research pipelines

For applicants in Nigeria

• Location is listed as Global and fully remote; preference for San Francisco Bay Area but remote work is supported
• The listing states fully remote work with travel twice per quarter and possible additional travel to DC and San Francisco
• The listing provides salary ranges in USD; confirm currency handling and any regional pay considerations before interviewing
• The listing does not specify timezone requirements beyond remote and travel expectations; confirm during interview
• The listing does not specify visa or local work authorization notes; confirm eligibility during interview