Software Engineer (Model Quality), Claude CodeSan Francisco, CA or New York, NY
About Anthropic
Anthropics 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.
About the Role
Were looking for a Software Engineer to work at the intersection of engineering and research on the Claude Code team. In this role, youll collaborate directly with Anthropics researchers to improve Claudes coding capabilities through tooling, infrastructure, and evaluations. Youll build systems that help us understand where Claude Code excels and where it falls short-and then help close those gaps.
Were looking for engineers who can build robust, complex systems and who thrive in fast-paced, high-intensity environments. Youll take ambiguous problems and turn them into reliable infrastructure that accelerates our research.
Responsibilities
Design and build eval systems that measure model capabilities across diverse coding tasks Build tooling and infrastructure that enables researchers to run experiments at scale Develop pipelines for data collection, processing, and analysis Create internal tools that improve researcher productivity and accelerate iteration cycles Serve as a bridge between product and research-bring strong product intuition to inform which capabilities matter most Work closely with researchers to translate research questions into engineering solutions Own systems end-to-end-from design through production reliability
You may be a good fit if you:
Have built and owned complex systems-pipelines, infrastructure, or software that orchestrates many components and handles significant state and logic Thrive in high-intensity environments with fast iteration cycles Take full ownership of problems and drive them to completion independently Are a power user of agentic coding tools and have strong intuition about model capabilities and limitations Are comfortable diving into unfamiliar technical domains and figuring things out quickly Care deeply about correctness and reliability in the systems you build Are excited to work at the boundary between engineering and AI research Have at least 5 years of work experience
Strong candidates may also have experience with:
Writing or maintaining eval frameworks Reinforcement learning systems Working in high-performance, demanding environments-trading firms, quant funds, competitive research labs, or fast-moving startups where intensity is the norm Have research computing or scientific infrastructure background Have a strong quantitative foundation (math, physics, or related fields) Python and TypeScript