Fullstack Software Engineer (Reinforcement Learning, AI)

Location: San Francisco, CA or New York City, NY (hybrid policy: at least 25% time in office)

Salary: $300,000 - $405,000 USD

Company

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems like Claude.

What you will do

• Build and extend web platforms for RL environment creation, management, and quality review including configuration, versioning, and validation workflows
• Develop vendor-facing interfaces and tooling for external partners to create and iterate on training environments
• Design platforms for scalable human data collection with labeling workflows, quality assurance, and feedback mechanisms
• Create evaluation dashboards and observability UIs for real-time insights into environment quality and training health
• Build backend services and APIs connecting environment tools, data systems, and RL training infrastructure
• Expand code data generation pipelines producing diverse programming tasks with robust reward signals

Requirements

• Strong software engineering fundamentals and fullstack experience from database to frontend
• Proficient in Python and modern web stack (React, TypeScript or similar)
• Track record of shipping systems solving hard problems with high impact
• High agency, clear communication, and ability to thrive in fast-moving environment
• Location: Must spend at least 25% time in San Francisco or New York office
• Bachelor’s degree or equivalent in relevant field

Nice to have

• Experience building data collection, labeling, or annotation platforms at scale
• Multi-tenant platforms with role-based access and vendor management
• Cloud infrastructure (GCP/AWS), Docker, CI/CD, async Python, LLM workflows
• Working with external vendors on technical integrations

Culture & Benefits

• Competitive compensation, equity, generous vacation and parental leave
• Flexible working hours and visa sponsorship (with effort to support)
• Collaborative team focused on high-impact AI research
• Location-based hybrid policy with office collaboration