North AmericaFull TimeEngineering
Remotely
cybersecurityautomated testingmodel evaluationagentic harnessesprompt injectionjailbreaking
Job Description
📋 Description
- Design and run rigorous evaluations of model cyber capabilities and safeguards.
- Hands-on testing to understand model capabilities for experienced security practitioners.
- Distinguish real-world risk from policy failures.
- Build automated testing infrastructure for repeatable measurement and rapid iteration.
- Test novel abuse risks in agentic systems (e.g., prompt injection, agent hijacking).
- Translate findings into risk assessments and recommendations for partners.
🎯 Requirements
- Strong background in cybersecurity or AI model evaluation.
- Experience building testing tools and automating experiments.
- Ability to communicate findings clearly to technical and non-technical audiences.
- Experience collaborating across Security, Research, Product, Policy, and Engineering teams.
- Willingness to contribute to Safety Bug Bounty work.
🎁 Benefits
- Hybrid work model with relocation assistance.