Anthropic
Research Engineer, Computer Use
$500K–$850K
San Francisco, CA +5 · Full-time · Posted Jun 30, 2026
- Location
- San Francisco, CA
New York City, NY
Seattle, WA
New York, New York, United States
San Francisco, California, United States
Seattle, Washington, United States - Experience
- Minimum years of experience correlate with internal job level requirements; no specific number stated.
- Education
- Bachelor’s degree or an equivalent combination of education, training, and/or experience; relevant field of study demonstrated through coursework, training, or professional experience.
- Visa sponsorship
- Visa sponsorship is available, but not guaranteed for every role or candidate; the company will make every reasonable effort to obtain a visa after an offer.
About the job
Advance Claude’s ability to safely and reliably perceive and operate real software by conducting research, developing evaluations and training environments, and bringing improvements into production.
What you'll do
- Design experiments and evaluation frameworks to improve and measure computer-use capabilities.
- Build computer-use and vision reinforcement learning environments, pipelines, and validation tools.
- Collaborate with model training, infrastructure, and product teams to improve training and deploy research advances.
What they ask for
- Software engineering experience and proficiency in Python.
- Experience training, fine-tuning, or evaluating machine learning models.
- Strong communication skills and a collaborative working style.
- Care about the societal impacts and safety of your work.
Nice to have
- Experience with computer-use or agentic models, reinforcement learning, multimodal training, or agent evaluations.
- Experience building reinforcement learning environments, simulation systems, or large-scale ML infrastructure, or working with product teams on model improvements.
Pay
$500,000 — $850,000 USD (base salary)
Tools
Python, Claude
AI focus
Research and engineering to improve Claude’s computer perception and agentic capabilities, including reinforcement learning environments, evaluations, and safe operation of software.
Model-written summary · the original posting is authoritative