Software Engineer, Infrastructure, Interpretability

When you see what modern language models are capable of, do you wonder, "How do these things work? How can we trust them?" The Interpretability team at Anthropic works to understand what is actually happening inside trained models and applies our best techniques to keep frontier AI safe as it rapidly improves. Think of us as doing "neuroscience" of neural networks using "microscopes" we build, or reverse-engineering neural networks like binary programs. This role is an early hire on a new infrastructure effort within Interpretability: you will help define its charter, not just execute it. You will deeply embed alongside Interp Researchers to understand their workflows, building your understanding of the research as you go while bridging communication with Anthropic's wider platform and security teams. Every hour of researcher friction you remove is multiplied across the whole organization, and the infrastructure you build sets the pace at which interpretability results reach real safety decisions.

Location: USA

- Architect and implement secure-by-default research environments and access patterns that enable deep model access while improving both security posture and research productivity.

- Build data-access patterns and guardrails that ensure policy adherence as interpretability work moves from theory into practical application across the organization.

- Manage research data at petabyte scale and make efficient use of large accelerator fleets, including storage lifecycle, capacity planning, and scheduling.

- Design and own shared infrastructure for Interpretability, including research environments, data systems, and core platform components that become the foundation for future work.

- Partner closely with interpretability researchers to understand their workflows, turning nuanced research requirements into robust infrastructure primitives and tooling.

- Remove friction for researchers by automating environment setup, data access, and compute orchestration so they can focus on insight rather than infrastructure.

- Collaborate with Anthropic's wider platform and security teams to align on standards, integrate services, and ensure long-term maintainability and scalability.

- Instrument and observe infrastructure to support rapid iteration, debugging, and reproducibility of interpretability experiments across model scales.

- Contribute to the design of agentic engineering tooling, observability, and developer experience that keeps research teams moving fast without compromising safety or compliance.

- Define and maintain interfaces between security, privacy, data management, and compute scheduling layers to ensure cohesive system behavior.

- Own end-to-end reliability and performance of infrastructure, driving incident response, capacity planning, and continuous improvement.

- Work cross-functionally to translate research roadmaps into infrastructure milestones, anticipating future needs as interpretability methods scale.

- Partner with security engineers to design access controls, audit trails, and data handling patterns that meet evolving policy requirements.

- Help establish best practices and documentation so that infrastructure serves as a stable platform for many concurrent interpretability initiatives.

- You are excited to build infrastructure that supports cutting-edge interpretability research at scale, understanding that this role directly impacts the safety and reliability of frontier AI systems.

- You have experience designing and operating large-scale distributed systems, with deep knowledge of storage, compute scheduling, and data lifecycle management.

- You are comfortable working in fast-paced research environments where requirements evolve quickly and you must turn ambiguous problems into robust infrastructure solutions.

- You have strong programming skills in systems-level languages and are comfortable building tools that researchers will depend on for their daily work.

- You understand the security and privacy implications of running large models and have experience implementing patterns that enforce policy by design.

- You are comfortable with petabyte-scale data concepts and have hands-on experience managing storage and compute at scale in cloud or high-performance computing environments.

- You have a track record of building developer-facing infrastructure and care deeply about usability, observability, and performance.

- You are comfortable collaborating with research teams to translate abstract requirements into concrete infrastructure contracts and guarantees.

- Experience with interpretability tooling, experiment tracking, or data versioning in ML contexts.

- Familiarity with model serving, inference-time instrumentation, or runtime security for AI workloads.

- Contributions to open-source infrastructure projects that align with Anthropic's goals.

- Hours, travel, visa, or deadlines are as specified in the source posting.

This role offers the opportunity to define the infrastructure backbone that keeps interpretability research secure, private, and frictionless, directly shaping the safety and scalability of Anthropic's AI systems.