Staff Software Engineer, Environments Infrastructure

Details

Anthropic’s Environments organization builds and maintains infrastructure that improves Claude’s capabilities through reinforcement learning. This includes frameworks for building environments and the infrastructure responsible for running them. The team’s mission is to productionize research by designing frameworks and APIs that help research teams work faster while keeping production reinforcement learning runs healthy, maintainable, monitored, and easy to triage.

The role requires deep Python expertise, strong API and framework design judgment, and an understanding of how complex systems fail, especially silently. Experience building and operating stateful distributed systems such as workflow engines, actor frameworks, or durable-execution runtimes is beneficial. The successful candidate should be comfortable working in research codebases, improving abstractions incrementally, and using AI tools with thorough verification.

Responsibilities

• Design widely used APIs, frameworks, and abstractions for engineers and researchers, making correct usage the default and structurally preventing classes of errors.
• Own the platform layers beneath every environment, including the agent runtime.
• Build tooling that enables environment owners to understand, debug, and maintain production environments without infrastructure-engineering support.
• Embed with research teams on a rotational basis, work directly in their codebases, and transfer ownership when rotating off.
• Anticipate silent failure modes and prevent them through type safety, well-designed invariants, targeted testing, and refactoring.
• Drive adoption of new frameworks across the organization, including deprecations and cutovers.
• Help define engineering standards, review practices, and design patterns for a new team.
• Mentor researchers and engineers in adopting engineering standards and frameworks.

Requirements

• Deep expertise in Python, including static typing, safe asynchronous and concurrency patterns, and performance-oriented programming.
• Strong API and framework design skills, with the ability to explain why an interface is appropriate and a track record of creating frameworks adopted by other engineers or teams.
• Experience designing or operating stateful concurrent or distributed systems.
• Ability to reason carefully about failure, retries, idempotency, and consistency.
• A verification-oriented approach, including measuring before drawing conclusions and building checks that demonstrate system correctness.
• Experience working productively in large, evolving, or research-style codebases that were not originally written by you.
• Strong written and verbal communication skills and comfort working through ambiguity.
• Bachelor’s degree or an equivalent combination of education, training, and experience. The field of study must be relevant to the role through coursework, training, or professional experience.

Preferred Qualifications

• Experience building infrastructure, tooling, or frameworks for machine learning research or reinforcement learning workflows.
• Familiarity with agentic systems or large language model training pipelines.
• Experience building agent frameworks, orchestration engines, or multi-agent systems, including checkpoint and restore, replay, and coordination of long-running stateful processes.
• Experience using AI coding tools on correctness-critical code, with good judgment about delegation and verification.
• Experience building client libraries or SDKs for sandboxed, containerized, or remote execution platforms.
• Experience with large-scale data processing, dataset lifecycle management, or data lineage systems.
• Experience designing serialization schemes, plugin systems, or extensible class hierarchies used across an organization.
• Experience embedding with or consulting for other teams and handing off systems for others to own.
• Experience defining code standards adopted across teams or serving as a technical lead.

Representative Projects

• Design a base reinforcement learning environment abstraction that supports most environments built across RL.
• Redesign the model-tool interface for sandboxed agentic environments so state survives serialization and tools cannot silently lose state.
• Design state-sharing and recovery models for multi-agent workloads so losing a sandbox during a task results in a transparent resume rather than lost work.
• Define failure and retry models for a sandboxed execution platform, distinguishing infrastructure faults from genuine task outcomes.
• Build tooling that allows environment owners to diagnose and resolve unhealthy production runs themselves.

Benefits

Anthropic offers competitive compensation and benefits, optional equity donation matching, generous vacation and parental leave, flexible working hours, and an office environment for collaboration. Anthropic also sponsors visas when possible and retains an immigration lawyer to assist with visa processes.