Software Engineer, RL Data

About this role

Anthropic seeks a senior engineer to build and own the infrastructure for collecting high-quality reinforcement learning training data for Claude. You'll design production pipelines, human feedback tooling, execution environments, and quality systems that power AI safety and capability research at scale.

What you'll do

• Design and own end-to-end data collection pipelines from prototype to production serving multiple research teams
• Build and iterate on prompts, evaluations, graders, and QA frameworks to ensure training data quality
• Develop sandboxed execution environments for complex, multi-step agentic tasks that scale to millions of rollouts
• Create user-friendly interfaces for human feedback collection and validation workflows
• Partner with research teams, operations, security, and compliance to roll out new data sources and systems
• Identify and eliminate bottlenecks in the task-to-training pipeline

What they're looking for

• End-to-end project ownership in ambiguous, fast-paced environments
• Software engineering in Python and/or TypeScript
• Data pipeline design and production data engineering
• Docker, Kubernetes, and cloud infrastructure
• RL concepts including evals, reward systems, and grading
• Cross-functional collaboration and stakeholder management
• AI tool proficiency for daily work
• Reinforcement learning on LLMs (preferred)

Likely interview questions

• Tell us about a major project you owned end-to-end in an ambiguous environment — what made it succeed and what would you do differently?
• Describe your experience building or operating data pipelines at scale. How have you approached data quality and reliability?