Anthropic
Staff+ Software Engineer, Claude Science
$405K–$485K
San Francisco, CA +2 · Staff · Full-time · Posted Sep 3, 2026
- Location
- San Francisco, CA
New York City, NY
San Francisco, California, United States - Experience
- 8+ years of software engineering experience; ideally 2+ years at Staff or equivalent technical leadership level. The posting also states that required years correlate with internal job-level requirements.
- Education
- Bachelor’s degree or an equivalent combination of education, training, and/or experience; relevant field demonstrated through coursework, training, or professional experience.
- Visa sponsorship
- Visa sponsorship is available, but not for every role or candidate; Anthropic says it will make every reasonable effort to obtain a visa after making an offer.
About the job
Leads product and technical development of Claude Science, an AI workbench that helps researchers conduct scientific work from discovery through publication.
What you'll do
- Shape product and architecture while shipping against a roadmap.
- Work with scientists and product/design teams to translate research workflows into product priorities and features.
- Partner with research teams to improve scientific model capabilities through evaluations, failure analysis, and user feedback.
What they ask for
- 8+ years of software engineering experience; ideally 2+ years at Staff or equivalent technical leadership level.
- Experience building products from 0 to 1 in fast-moving environments and setting technical direction with limited precedent.
- Experience building AI products and turning model capabilities into useful applications.
- Ability to work directly with technical domain experts and drive cross-team alignment.
Nice to have
- Background in chemistry, biology, physics, or another science.
- Experience improving domain-specific model capabilities with research teams, including evaluation frameworks.
Pay
$405,000 — $485,000 USD (base salary)
AI focus
Build an AI workbench for scientific research and help improve models' scientific capabilities through evaluations, failure analysis, and real-world user feedback.
Model-written summary · the original posting is authoritative