Staff Software Engineer, Environments Infrastructure

About Anthropic Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems. About the role Anthropic's Environments organization builds and maintains the infrastructure that improves Claude’s capabilities through reinforcement learning. That includes the frameworks researchers use to build environments and the infrastructure responsible for running them. The team's mission is to productionize research. You'll embed with research teams, get up to speed on how they work, and design the frameworks and APIs that let them move faster, building systems the team can understand, own, and maintain themselves. Scope also includes keeping production RL runs healthy, maintainable, monitored, and easy to triage. You'll be a strong fit if you have deep expertise in Python, a refined sense of taste for API and framework design, and good intuition for how complex systems fail, especially silently. It's a bonus if you've built and operated a stateful distributed system, such as a workflow engine, actor framework, or durable-execution runtime, where correctness depends on getting shared state and recovery right. You should be comfortable diving into messy research code, finding the abstractions that matter, and improving them incrementally while researchers continue to build on your work. You should also be comfortable using AI tools to accelerate your own development, but have an impulse towards deep verification. Key responsibilities • Design widely used APIs, frameworks, and abstractions that other engineers and researchers build on, making correct usage the default and ruling out entire classes of errors structurally • Own the platform layers that sit beneath every environment, including the agent runtime • Build the tooling that lets environment owners understand, debug, and maintain their environments in production without needing an infrastructure engineer in the loop • Embed with research teams on a rotational basis, work directly in their codebases without slowing down the research they support, and transfer ownership when you rotate off • Anticipate silent failure modes and prevent them structurally through type safety, well-designed invariants, targeted testing, and refactors that reduce the room for correctness issues • Drive adoption of new frameworks across the organization, including deprecations and cutovers • Help define the engineering standards, review practices, and design patterns for a new team, and mentor researchers and engineers in adopting them Minimum qualifications • Deep expertise in Python, including static typing, safe async and concurrency patterns, and writing performant code • Strong taste in API and framework design, the ability to explain why an interface is right or wrong rather than just recognizing it, and a track record of other engineers or teams adopting and building on frameworks you have built • Experience designing or operating stateful concurrent or distributed systems, and reasoning carefully about failure, retires, idempotency, and consistency • A habit of verification: you measure before you conclude, and you build the checks that let a system show it's correct • Experience working productively in large, evolving, or research-style codebases that you didn't originally write • Strong written and verbal communication with collaborators of varied engineering backgrounds, and comfort with ambiguity: able to scope your own work from a loosely defined problem and drive it to a maintainable outcome Preferred qualifications • Experience building infrastructure, tooling, or frameworks for machine learning research or RL workflows, and familiarity with agentic systems or LLM 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 code where correctness matters, with good judgment about what to delegate and how to make the results verifiable • Experience building client libraries or SDKs on top of 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, or defining code standards adopted across teams, or prior experience as a technical lead Representative projects These are examples of the challenges the team tackles: • Design a base RL environment abstraction that can be subclassed to support the large majority of environments built across RL • Redesign the model-tool interface for sandboxed agentic environments so that state is guaranteed to survive serialization, making it structurally impossible to write a tool that silently loses state • Design the state-sharing and recovery model for multi-agent workloads, so that losing a sandbox partway through a task becomes a transparent resume rather than lost work • Define the failure and retry model for a sandboxed execution platform, distinguishing infrastructure faults from genuine task outcomes so that each is handled correctly • Build the tooling that lets an environment owner diagnose why their environment is unhealthy in a production run, and fix it themselves The annual compensation range for this role is listed below. For sales roles, the range provided is the role’s On Target Earnings ("OTE") range, meaning that the range includes both the sales commissions/sales bonuses target and annual base salary for the role. Annual Salary: $405,000—$605,000 USD Logistics Minimum education: Bachelor’s degree or an equivalent combination of education, training, and/or experience Required field of study: A field relevant to the role as demonstrated through coursework, training, or professional experience Minimum years of experience: Years of experience required will correlate with the internal job level requirements for the position Location-based hybrid policy: Currently, we expect all staff to be in one of our offices at least 25% of the time. However, some roles may require more time in our offices. Visa sponsorship: We do sponsor visas! However, we aren't able to successfully sponsor visas for every role and every candidate. But if we make you an offer, we will make every reasonable effort to get you a visa, and we retain an immigration lawyer to help with this. We encourage you to apply even if you do not believe you meet every single qualification. Not all strong candidates will meet every single qualification as listed. Research shows that people who identify as being from underrepresented groups are more prone to experiencing imposter syndrome and doubting the strength of their candidacy, so we urge you not to exclude yourself prematurely and to submit an application if you're interested in this work. We think AI systems like the ones we're building have enormous social and ethical implications. We think this makes representation even more important, and we strive to include a range of diverse perspectives on our team. Your safety matters to us. To protect yourself from potential scams, remember that Anthropic recruiters only contact you from @anthropic.com email addresses. In some cases, we may partner with vetted recruiting agencies who will identify themselves as working on behalf of Anthropic. Be cautious of emails from other domains. Legitimate Anthropic recruiters will never ask for money, fees, or banking information before your first day. If you're ever unsure about a communication, don't click any links—visit anthropic.com/careers directly for confirmed position openings. How we're different We believe that the highest-impact AI research will be big science. At Anthropic we work as a single cohesive team on just a few large-scale research efforts. And we value impact — advancing our long-term goals of steerable, trustworthy AI — rather than work on smaller and more specific puzzles. We view AI research as an empirical science, which has as much in common with physics and biology as with traditional efforts in computer science. We're an extremely collaborative group, and we host frequent research discussions to ensure that we are pursuing the highest-impact work at any given time. As such, we greatly value communication skills. The easiest way to understand our research directions is to read our recent research. This research continues many of the directions our team worked on prior to Anthropic, including: GPT-3, Circuit-Based Interpretability, Multimodal Neurons, Scaling Laws, AI & Compute, Concrete Problems in AI Safety, and Learning from Human Preferences. Come work with us! Anthropic is a public benefit corporation headquartered in San Francisco. We offer competitive compensation and benefits, optional equity donation matching, generous vacation and parental leave, flexible working hours, and a lovely office space in which to collaborate with colleagues. Guidance on Candidates' AI Usage: Learn about our policy for using AI in our application process.