Research Engineer, Frontier Red Team (RSP Evaluations)

Job Description

Anthropic is seeking a Research Engineer to join their Frontier Red Team, focusing on Responsible Scaling Policy (RSP) evaluations. This role involves developing and running evaluations for catastrophic risks, ensuring the safe deployment of AI models. The Research Engineer will collaborate with domain experts across biosecurity, autonomous replication, cybersecurity, and national security to measure dangerous capabilities in models and determine if they cross ASL thresholds.

The role involves:

• Designing and implementing robust evaluation infrastructure.
• Leading technical projects to build and scale evaluation systems.
• Collaborating with domain experts to translate insights into evaluation frameworks.
• Building sandboxed testing environments and automated pipelines.
• Partnering with cross-functional teams to advance Anthropic's safety mission.
• Contributing to Capability Reports.

Requirements include:

• Experience with fast, iterative experiments with frontier AI models.
• Experience designing or implementing evaluations involving sampling + prompting LLMs.
• Strong software engineering skills with Python experience.
• Experience working with distributed systems.
• Comfort defining technical specifications and executing towards them.
• Ability to thrive in fast-paced, collaborative environments.
• Care deeply about AI safety and responsible development.

Anthropic offers:

• Competitive compensation and benefits.
• Optional equity donation matching.
• Generous vacation and parental leave.
• Flexible working hours.