SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Anthropic's Takeoff Intel team measures how AI systems are increasingly delegating AI development tasks to themselves—a form of recursive self-improvement. As a Research Engineer, you'll build and operate the evaluation and measurement infrastructure that makes this research possible.
You'll work on a small, generalist team spanning evaluation infrastructure, large-scale data processing, and analysis tooling. The role emphasizes shipping working prototypes quickly rather than building perfect systems. You'll design and run capability evaluations at scale, build data and analysis pipelines that convert model outputs and telemetry into reliable metrics, prototype new measurement instruments rapidly, and validate what's worth keeping. You'll also review AI-written code as part of normal workflow and collaborate closely with research scientists and partner teams to define what matters to measure.
Your work directly influences Anthropic's internal research priorities and safety planning, and contributes to public reporting on AI progress acceleration. Recent outputs include Anthropic's Epoch Capabilities Index adaptation (published in all recent system cards), AI R&D capability assessments in Claude system cards, and data underlying the "When AI Builds Itself" research.
Ideal candidates have shipped evaluations, data products, or research libraries end-to-end; prototype quickly and discard code without attachment; handle messy, large-volume data pragmatically; have run experiments on large language models (not just processed their outputs); work from vague research questions rather than detailed specs; and communicate results clearly to collaborating researchers. Strong additional experience includes building LLM evaluation harnesses or benchmark infrastructure, large-scale ML/data infrastructure (self-driving, observability), tools that other researchers depend on, or a track record of catching errors in AI-written code.
The team is hiring at both junior and senior levels. The role is remote-friendly with travel required and a primary office in San Francisco.
