Staff+ Software Engineer, Safeguards Review Tooling

Anthropic•San Francisco, CA

18h•Hybrid

About The Position

The Safeguards team at Anthropic is responsible for ensuring the safe development and deployment of AI models and products. The Review Tooling team, specifically, builds systems that human reviewers and Claude use to investigate potential harms and take enforcement actions across Anthropic's products and third-party cloud platforms. This is a foundational role where the engineer will own the tools for safety investigators, including analytics, privacy-preserving primitives, and a sandbox environment for developing new interfaces. The role also involves scaling review through automation, enabling Claude to assist human reviewers while keeping human judgment central where it matters most. These internal tools are critical for identifying harmful behavior, making enforcement decisions, and feeding signal back into model training. The engineer will collaborate closely with policy, operations, data science, legal, and privacy teams to ensure effective, accurate, and trustworthy enforcement systems.

Requirements

A technical background in full-stack or platform engineering, with the ability to engage deeply in architecture and design discussions
Experience shipping internal tools or platforms with demanding operational users, and a track record of improving their workflows measurably
Experience working cross-functionally with non-engineering partners such as operations, policy, or legal teams
Excellent communication skills, including the ability to explain technical tradeoffs to non-technical stakeholders
Care about the societal impacts of AI and want your work to make powerful systems safer

Nice To Haves

8+ years of industry software engineering experience
Experience building trust and safety, integrity, fraud, or abuse-prevention tooling, or other systems supporting human review at scale
Experience designing systems under strict privacy, compliance, or data governance constraints, such as zero data retention environments
Experience integrating LLMs or agentic systems into operational workflows, or building human-in-the-loop automation — including using agentic coding tools (e.g., Claude Code) as a core part of your own development workflow
Experience building developer platforms or extensible tooling frameworks that other teams build on top of
Experience supporting enforcement or moderation systems across multiple product surfaces, including enterprise or cloud platform contexts
A product-minded approach to internal users: you work directly with the people using your tools, watch where they struggle, and fix it

Responsibilities

Build investigation, review, and enforcement tooling for both first-party and third-party platform surfaces — including case queues, investigation views, decision and audit logging, and account-actioning workflows
Develop the platform layer of reusable APIs, data storage, and backend services that lets new review workflows be stood up quickly and safely
Scale review through automation, including enabling reviewers to use Claude effectively and building toward Claude-assisted and Claude-driven review workflows
Partner with policy, operations, legal, privacy, and data science stakeholders to translate enforcement and investigation needs into reliable, well-designed systems that measurably reduce handling time and decision error
Build in the guardrails that sensitive internal tools require: granular permissions, audit trails, data-access controls, and reviewer wellbeing features such as content obfuscation and exposure limits
Instrument the tools you ship — surfacing metrics on queue health, reviewer throughput, and decision quality — and ensure tooling evolves alongside new privacy primitives and data retention commitments