RL Frontier Engineer: Scale-Savvy AI Architect

Anthropic; New York (NY); Hybrid; USD 500,000 - 850,000; Full time

Benefits offered by this job

Competitive compensation

Generous vacation

Flexible working hours

Job summary

Anthropic is seeking an RL-focused engineer to bridge research and engineering efforts. You will build architectures and RL algorithms that scale from small experiments to frontier-scale runs, diagnosing differences that arise as scale increases and ensuring reliable, repeatable progress.

You will own end-to-end performance from research code to hardware, while developing cost models and infra to accelerate iteration on large-scale RL systems.

Qualifications

• Deep familiarity with transformer language models, training dynamics, and large-scale optimisation.
• Experience training large models in a distributed setting with data, tensor, and pipeline parallelism.
• A track record of original ML research or system work with measurable impact.
• Ability to design rigorous, scalable experiments with baselines and ablations.
• Ability to reason about compute, memory, and communication costs.
• Strong Python and PyTorch/JAX skills across the stack.

Responsibilities

• Study RL training and sampling scale with model size, context length, and compute.
• Develop next-generation model architectures and RL algorithms at frontier scale.
• Translate promising small-scale results to frontier-scale runs and diagnose discrepancies.
• Build experimental infrastructure for fast, reproducible comparisons.
• Own end-to-end performance of our largest RL runs, from research code to hardware.
• Build performance and cost models to guide architecture and algorithm changes.
• Investigate training dynamics at scale, including instabilities and throughput regressions.

Skills

Transformer models

Distributed training

Python & PyTorch

C++

Rust

ML research

Education

Bachelor's degree or equivalent

Tools

C++

Rust

Job description

Anthropic is seeking an RL-focused engineer to bridge research and engineering efforts. You will build architectures and RL algorithms that scale from small experiments to frontier-scale runs, diagnosing differences that arise as scale increases and ensuring reliable, repeatable progress.

You will own end-to-end performance from research code to hardware, while developing cost models and infra to accelerate iteration on large-scale RL systems.