Cognition
Head of Safety
San Francisco · On-site · Director · Full-time · Posted Oct 9, 2026
- Education
- An advanced degree in Computer Science, Machine Learning, or a related field is a plus.
About the job
Build and lead Cognition’s safety function for Devin and its underlying models, covering research, evaluation, training, deployment policy, and team leadership.
What you'll do
- Set safety strategy and own safety outcomes for Devin and its models.
- Create evaluations and red-teaming for agentic risks, and integrate them into model and product releases.
- Shape safety in training and the agent harness, set deployment policies, and hire and lead safety staff.
What they ask for
- Hands-on safety, alignment, evaluation, or red-teaming experience at a frontier AI lab, including shipped work affecting real models or products.
- Depth in agentic systems and ways to measure and mitigate their risks.
- Track record of published or widely used safety research, evaluations, or methods.
- Technical fluency with Python and the training and inference stack.
Nice to have
- Advanced degree in Computer Science, Machine Learning, or a related field.
Tools
Devin, Python
AI focus
Develop and implement safety and alignment for autonomous software agents, including risk evaluation and red-teaming, training and harness mitigations, and deployment policies.
Model-written summary · the original posting is authoritative