Anthropic
Research Engineer, RL Engineering
$500K–$850K
San Francisco, CA +3 · Full-time · Posted Oct 14, 2025
- Location
- San Francisco, CA
New York City, NY
Seattle, WA
San Francisco, California, United States - Experience
- Years required correlate with the internal job level; no specific number is stated.
- Education
- Bachelor’s degree or an equivalent combination of education, training, and/or experience; relevant field of study as demonstrated through coursework, training, or professional experience.
- Visa sponsorship
- Visa sponsorship is available, but not for every role or candidate; the company says it will make every reasonable effort to obtain a visa if it makes an offer.
About the job
Build and improve the reinforcement-learning training system used for Anthropic’s production models and research experiments, while studying and optimizing RL training at scale.
What you'll do
- Build, own, and improve the core RL training system across orchestration, environments, training, inference, and evaluation.
- Study RL training at scale and implement new training methods as stable, efficient, well-tested code.
- Profile and optimize training and evaluation, debug system failures, and make the system easier for researchers to use.
What they ask for
- Proficiency in Python and experience working in, debugging, and improving a large ML codebase.
- Experience with large-scale machine learning training or the systems that support it.
- Experience with at least one modern ML framework, such as JAX or PyTorch.
- Ability to design trustworthy controlled experiments and balance research exploration with engineering implementation.
Nice to have
- Experience with reinforcement learning for large language models or large-scale distributed training systems.
- Experience with Rust or C++.
Pay
USD 500,000–850,000 annual base salary
Tools
Python, JAX, PyTorch, Rust, C++
AI focus
Develop and improve RL training systems and methods for production and research models, supporting safe and beneficial AI systems.
Model-written summary · the original posting is authoritative