Software Engineer, Infrastructure, Interpretability

Details

Anthropic’s Interpretability team works to understand what happens inside trained models and applies those techniques to help keep frontier AI safe. This role is an early hire for a new infrastructure effort within Interpretability and will help define its charter while building secure, private, and low-friction access to frontier models for researchers.

The work spans security, privacy, data and compute management, and developer experience. You will work closely with Interpretability researchers while coordinating with platform and security teams across Anthropic.

Responsibilities

• Design, build, and own shared infrastructure for Interpretability, including research environments, data systems, and compute tooling.
• Lead cross-team efforts with agentic engineering, security, compute, and storage platform teams to ensure company-wide solutions serve research needs.
• Discover and resolve major organization-wide developer experience issues.
• Help take interpretability methods from research code to dependable audit pipelines.
• Design secure-by-default environments and access patterns for deep model access.
• Build privacy-focused data-access patterns that support policy adherence.
• Manage research data at petabyte scale, including storage lifecycle and capacity planning.
• Improve the efficiency of large accelerator fleets through scheduling and compute management.
• Build developer tooling and observability that help researchers work efficiently.

Requirements

• High proficiency in at least one programming language, such as Python, Rust, Go, or Java, and productivity with Python.
• Significant experience building and operating secure and scalable software infrastructure, including cloud systems, distributed systems, or developer tooling.
• Strong cross-functional communication skills and the ability to work with both researchers and platform and security teams.
• Curiosity about unfamiliar domains.
• Ability to prioritize impactful work, operate with ambiguity, and question assumptions.
• Interest in interpretability research and its role in AI safety; research experience is not required.
• Care for the societal impacts and ethics of the work.
• A bachelor’s degree or equivalent combination of education, training, and experience.
• A field of study relevant to the role, demonstrated through coursework, training, or professional experience.

Strong candidates may also have experience with GCP or AWS, Kubernetes, networking, infrastructure-as-code, identity and access management, sandboxing, red teaming, data warehousing, large-scale storage systems, data lifecycle management, compute schedulers, accelerator fleet management, developer productivity tooling, observability stacks, or tooling for research teams.

Representative Projects

• Stand up a hardened research environment for experimenting directly with frontier model weights.
• Build lifecycle management for petabytes of research data, including visibility, retention, and cost efficiency.
• Build self-service scheduling and capacity tooling.
• Create observability that catches infrastructure regressions before they cost researchers valuable time.

Benefits

Anthropic offers competitive compensation and benefits, optional equity donation matching, generous vacation and parental leave, flexible working hours, and office space for collaboration. Anthropic sponsors visas and retains an immigration lawyer to assist with the process, although sponsorship is not guaranteed for every role or candidate.

The role is based in the San Francisco office, with exceptional candidates potentially considered for remote work on a case-by-case basis. Anthropic currently expects staff to be in one of its offices at least 25% of the time, although some roles may require more office time.