Frontier-Scale RL Research Engineer

Anthropic Limited; Seattle (WA); Hybrid; USD 500,000 - 850,000; Full time

Job summary

Anthropic in Seattle is seeking a researcher-engineer to advance RL scaling across frontier-scale runs. You will design architectures and RL algorithms, push experiments from small to large scales, and optimize the systems that enable rapid iteration and trustworthy results.

You will own performance, develop cost models, and diagnose numerical or systemic issues that arise as models and compute scale dramatically. This work blends research and engineering.

Qualifications

• Deep familiarity with modern transformer language models and large-scale optimization.
• Hands-on experience training large models in distributed settings.
• Track record of original technical ML work with research, open-source, or production impact.
• Ability to design rigorous experiments at scale with baselines and ablations.
• Ability to reason about compute, memory, and communication costs.

Responsibilities

• Study RL training and sampling scale with model size and compute; identify scaling improvements.
• Develop next-generation model architectures and RL algorithms for frontier-scale runs.
• Translate small-scale results into frontier-scale experiments and diagnose failures.
• Build fast, reproducible experimental infrastructure for architecture comparisons.
• Own end-to-end performance of largest RL runs, from code to hardware.
• Create performance and cost models to guide architectural choices.
• Investigate training dynamics and instability sources at scale.

Skills

Transformer models

Distributed training

ML research training

Experiment design at scale

Python programming

JAX

PyTorch

System performance analysis

Cost modeling

Analytical reasoning

Education

Bachelor's degree or equivalent combination of education, training, and/or experience

Tools

C++

Rust

Job description

Anthropic in Seattle is seeking a researcher-engineer to advance RL scaling across frontier-scale runs. You will design architectures and RL algorithms, push experiments from small to large scales, and optimize the systems that enable rapid iteration and trustworthy results.

You will own performance, develop cost models, and diagnose numerical or systemic issues that arise as models and compute scale dramatically. This work blends research and engineering.