Staff Engineer, ML Evaluation & Infrastructure

Anthropic; York and North Yorkshire; On-site; GBP 100,000 - 140,000; Full time

Benefits offered by this job

Health insurance

Equity packages

Relocation support

Education stipend

Job summary

Anthropic is seeking a Staff Software Engineer for the Claude Code team in the United Kingdom. You will set technical direction at the intersection of engineering and research, architecting evaluation and research infrastructure and leading end-to-end ownership of critical systems.

You will mentor engineers, translate research questions into durable engineering solutions, and partner with research leads to shape capabilities and progress metrics while maintaining high standards of reliability

Qualifications

• Senior individual contributor with 10+ years of software engineering experience.
• Experience operating as a Staff or Principal engineer at a high-caliber organization.
• Strong correctness and reliability focus; raised engineering standards on teams.

Responsibilities

• Set technical direction for evaluation systems, research infrastructure, and internal tooling for the Claude Code team.
• Architect eval frameworks to measure model capabilities across coding tasks and scale with the roadmap.
• Lead design of infrastructure enabling researchers to run experiments at scale and make long-term tradeoffs.
• Mentor engineers, review designs, unblock peers, and model engineering standards at scale.
• Own critical systems end-to-end from architecture through production reliability.

Skills

Python

TypeScript

Engineering leadership

Job description

Anthropic is seeking a Staff Software Engineer for the Claude Code team in the United Kingdom. You will set technical direction at the intersection of engineering and research, architecting evaluation and research infrastructure and leading end-to-end ownership of critical systems.

You will mentor engineers, translate research questions into durable engineering solutions, and partner with research leads to shape capabilities and progress metrics while maintaining high standards of reliability