This role builds the evaluation infrastructure for Anthropic’s safety systems, designing experiments to measure investigative agent performance, creating high-quality eval datasets, and shipping methods into production pipelines. The successful candidate will work at the intersection of applied ML research and engineering, producing reliable metrics and tools to evaluate the safety of our AI models.
Staff+ Software Engineer, Safeguards Evals at Anthropic
Hybrid - San Francisco, CA | New York City, NY
More jobs at AnthropicSalary
USD 320,000 - 485,000
Requirements
Skills
- Proficiency in Python
- Experience building and maintaining data pipelines
- Experience working with LLMs and understanding of agentic systems with tool use and multi-step reasoning
- Strong data analysis skills
- Ability to move fluidly between research prototyping and production-quality code
- Ability to translate ambiguous problems into concrete, testable experiments
- 8+ years of industry software engineering experience
- Expertise in building or contributing to agent evaluation frameworks, benchmarks, or automated grading systems
- Extensive experience in trust and safety, content moderation, or abuse detection systems
- Experience in red teaming, adversarial testing, or jailbreak research on AI systems
- Experience with synthetic data generation or data augmentation
- Experience with distributed systems or large-scale data processing
- Experience with prompt engineering or building LLM-powered applications
- Bachelor’s degree or equivalent education, training, or experience
Responsibilities
- Build and own the evaluation harness for an agentic investigation system — defining metrics, test cases and grading approaches for a complex long horizon agent
- Construct high-quality eval datasets representing real-world misuse across harm areas (e.g., cyber attacks, bio weapons, influence operations), drawing from real traffic patterns and synthetic generation
- Measure agent performance end-to-end (detection precision/recall, investigation quality, robustness) and drive hill-climbing on the hardest harm areas
- Analyze coverage to identify measurement gaps, and evolve evals so they remain unsaturated and high-signal as agent capabilities advance
- Productionize successful research into regression and release pipelines that run on every agent change, prompt update, and underlying model upgrade
- Build tooling that enables policy experts to author, run, and iterate on evaluations without engineering support
- Construct RL environments to improve Claude’s safety investigation capabilities
Technologies
PythonLLMsRL environmentsSynthetic data generationLarge-scale data processingAgent evaluation frameworks
See if your resume is ready for this job
See how our AI can optimize your resume and improve your chances for this role.