Research Scientist, Takeoff Intel

Anthropic's mission is to create reliable, interpretable, and steerable AI systems that are safe and beneficial for users and society.

We're looking for a Research Scientist to focus on measuring and understanding recursive-self-improvement in large models. You will use your judgment to decide what's worth measuring, design evaluations and models, and interpret results.

Responsibilities:

• Identify signals that track AI R&D acceleration and design evaluations to measure them
• Build quantitative models of capability growth and self-improvement dynamics
• Run experiments and evals to test hypotheses about automation and capability
• Make opinionated research bets and own the outcome
• Write graded assessments of what measurements show
• Collaborate with pretraining, RL, economic research, and policy teams

Requirements:

• Hands-on research experience with large language models
• Strong quantitative instincts and comfort with quantitative modeling
• Experience in forecasting and publishing AI forecasting scenarios
• Ability to design evaluations and defend methodology
• Clear writing and calibration skills
• Motivation by impact and comfort with graded assessments
• Care about AI safety and rapid capability growth

Strong candidates may also have:

• Trained or RL'd frontier models hands-on
• Experience with scaling laws, capability forecasting, or emergent-capability studies
• A physics, applied-math, or similarly quantitative background
• Written system card sections, capability reports, or methodology documents
• Experience supervising and correcting AI-written code

The annual compensation range for this role is $350,000-$850,000 USD.

Logistics:

• Minimum education: Bachelor's degree or equivalent
• Required field of study: Relevant to the role
• Minimum years of experience: Correlates with internal job level requirements
• Location-based hybrid policy: 25% office time
• Visa sponsorship: Available