Research Scientist, Frontier Red Team (Autonomy)

Anthropic Pbc

295d • $280,000 - $425,000

About The Position

Anthropic's mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems. We are looking for Research Scientists to develop and productionize advanced autonomy evaluations on our Frontier Red Team. Our goal is to develop and implement a gold standard of advanced autonomy evals to determine the AI Safety Level (ASL) of our models. This will have major implications for the way we train, deploy, and secure our models, as detailed in our Responsible Scaling Policy (RSP). We believe that developing autonomy evals is one of the best ways to study increasingly capable and agentic models. If you've thought particularly hard about how models might be agentic and associated risks, and you've built an eval or experiment around it, we'd like to meet you.

Requirements

• Have an ML background and experience leading experimental research on LLMs/multimodal models and/or agents.
• Have strong Python-based engineering skills.
• Are driven to find solutions to ambiguously scoped problems.
• Design and run experiments and iterate quickly to solve machine learning problems.
• Thrive in a collaborative environment.

Nice To Haves

• Experience training, working with, and prompting models.

Responsibilities

• Lead the end-to-end development of autonomy evals and research, starting with risk and capability modeling.
• Design, implement, and regularly run autonomy evaluations.
• Quickly iterate on experiments to evaluate autonomous capabilities and forecast future capabilities.
• Provide technical leadership to Research Engineers to scope and build scalable and secure infrastructure for large-scale experiments.
• Communicate the outcomes of the evaluations to relevant Anthropic teams, policy stakeholders, and research collaborators.
• Collaborate with other projects on the Frontier Red Team, Alignment, and beyond to improve infrastructure and design safety techniques for autonomous capabilities.

Benefits

• Visa sponsorship available.
• Location-based hybrid policy: expected to be in the office at least 25% of the time.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level
• Education Level: Bachelor's degree