320K–485K USD / year
Staff+ Software Engineer, Safeguards Evals
AIHybrid — New York City, NY, San Francisco, CA, Seattle, WA
Published on 2026-11-09
These details were extracted automatically from the original listing. They may not be complete or fully up to date — it's worth checking the original job post.
About this role
Build evaluation methods and infrastructure to measure model safety and the performance of agentic systems that investigate potential misuse. Develop representative datasets, validate evaluation quality, and integrate successful methods into training, regression, and release pipelines.
About the company
Anthropic builds reliable and interpretable AI systems, including the Claude family of models, for businesses and developers. The company focuses on AI safety research and practical AI assistants for enterprise and consumer use.
What you'll do
- Design experiments to improve evaluation quality, coverage, and grading accuracy
- Build and own evaluation harnesses for long-horizon agentic investigations
- Create datasets representing real-world misuse, using real traffic patterns and synthetic generation
- Build reinforcement learning environments to improve safety investigations
- Productionize evaluations in training, regression, and release pipelines
- Develop tools for policy experts to author and run evaluations
- Share evaluation findings with research and training teams
What we're looking for
- Experience building and maintaining data pipelines
- Experience with LLMs and understanding of their capabilities and failure modes, especially tool-using agents
- Strong data analysis skills
- Ability to move between research prototyping and production-quality code
- Ability to turn ambiguous problems into concrete, testable experiments
- Care about AI safety and its real-world impact
Nice to have
- 8+ years of software engineering or ML engineering experience
- Experience with LLM or agent evaluation frameworks, benchmarks, or automated grading
- Trust and safety, content moderation, or abuse detection experience
- Experience in AI red teaming, adversarial testing, or jailbreak research
- Synthetic data generation or data augmentation experience
- Distributed systems or large-scale data processing experience
- Prompt engineering or LLM-powered application development
Benefits
- Generous vacation and parental leave
- Flexible working hours
- Optional equity donation matching