Staff Software Engineer: RL Frameworks & APIs

Anthropic Limited; Seattle (WA); Hybrid; USD 405,000 - 625,000; Full time

Job summary

Anthropic Limited is seeking an accomplished software engineer to advance reinforcement learning tooling and agentic coding environments. You will design APIs and frameworks used by researchers, embed with teams on rotation, and shape conventions that keep systems maintainable as the organization grows.

You'll work primarily in Python, focus on reliable, well-typed code, and improve production RL pipelines.

Qualifications

• Deep expertise in Python, including typing, async and performant code.
• Track record designing intuitive, safe APIs or frameworks used by teams.
• Experience contributing to large, evolving research-style codebases.
• Ability to anticipate silent failure modes and prevent them via design and testing.
• Strong written and verbal communication across varied engineering backgrounds.
• Able to define scope in ambiguity and deliver maintainable outcomes.

Responsibilities

• Design widely-used APIs, frameworks, and abstractions for researchers and engineers.
• Embed with research teams on a rotational basis, transferring ownership.
• Work in research codebases to improve reliability without slowing research.
• Anticipate silent failures and prevent them through type safety and testing.
• Contribute to production RL system reliability, including monitoring and triage tooling.
• Define engineering standards and mentor researchers and engineers.

Skills

Python expertise

API/framework design

Large codebases experience

Failure mode anticipation

Communication skills

Ambiguity handling

Education

Bachelor's degree

Job description

Anthropic Limited is seeking an accomplished software engineer to advance reinforcement learning tooling and agentic coding environments. You will design APIs and frameworks used by researchers, embed with teams on rotation, and shape conventions that keep systems maintainable as the organization grows.

You'll work primarily in Python, focus on reliable, well-typed code, and improve production RL pipelines.