Frontier RL Systems Engineer & Research

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

Job summary

Anthropic is seeking a Reinforcement Learning Engineer to join the RL Scaling team. You will develop next-generation architectures and RL algorithms, and take them from small results to frontier-scale runs while ensuring efficient iteration across growing compute and models.

You will own end-to-end performance of our largest RL runs, diagnose numerical or systemic issues, and build performance and cost models to guide scalable research investments.

Qualifications

• Deep familiarity with modern transformer language models and training dynamics.
• Hands-on experience training large models in a distributed setting.
• Track record of original ML research or production impact.
• Ability to design rigorous experiments at scale with baselines and ablations.
• Ability to reason about compute, memory, and communication costs.
• Strong Python programming and experience with JAX or PyTorch.

Responsibilities

• Study RL training and sampling scaling with model size and compute.
• Develop next-generation model architectures and RL algorithms for frontier scale.
• Take small-scale results to frontier-scale runs and diagnose variances.
• Build infrastructure for fast, reproducible comparisons at meaningful scale.
• Own end-to-end performance of the largest RL runs across hardware.
• Build models to estimate throughput and cost of architecture changes.
• Investigate training dynamics and root causes of instabilities at scale.

Skills

Transformer models

Distributed training

Python

JAX or PyTorch

Experiment design

Compute cost awareness

Education

Bachelor’s degree

Tools

C++

Rust

Job description

Anthropic is seeking a Reinforcement Learning Engineer to join the RL Scaling team. You will develop next-generation architectures and RL algorithms, and take them from small results to frontier-scale runs while ensuring efficient iteration across growing compute and models.

You will own end-to-end performance of our largest RL runs, diagnose numerical or systemic issues, and build performance and cost models to guide scalable research investments.