Jobs at Epoch AI

Design and maintain a realistic benchmark suite of open-ended office tasks, create grading rubrics, run and analyze frontier AI models, produce public-facing evaluations and visualizations, and improve/automate evaluation workflows.

Source and engage high-context AI researchers, engineers, and operations candidates by mapping talent landscapes, running outreach and events, building referral networks, and maintaining a clean sourcing pipeline (Lever) while partnering with hiring managers to translate needs into hireable profiles.

Run, maintain, and expand an AI benchmarking infrastructure: implement and integrate benchmarks (primarily using the Inspect library), develop new benchmarks and prototypes, collaborate with researchers and engineers, and ensure evaluation outputs are accurate and integrated into research products.

Own and run the organization’s events and live activities end-to-end: plan multi-day staff retreats, programmatic events, and podcast recordings. Manage vendors, budgets, travel, RSVPs, on-site logistics, and design attendee experience. Build repeatable processes, decide when to outsource, and attend events in person. Role is Bay Area–based with some travel and may manage contractors as function grows.

Open call for expressions of interest across research, operations, engineering, data, and communications. Epoch AI is fully remote, hires in many countries, prefers full-time, and may require meetings during UTC-8 to UTC+1. Submit interest to be contacted for suitable openings.

Lead and conduct research projects analyzing AI capabilities, datacenters, supply chains, benchmarks, and trends. Produce datasets, benchmarks, reports, and forecasts; evaluate technical claims; collaborate across teams and communicate results clearly. Drive new research directions and prototype evaluation methods.

Contract data scientist to remotely gather, extract, and analyze technical information about AI models, benchmarks, data centers, and companies. Tasks include reading ML papers, tracking leaderboards, processing satellite images/permits, compiling infrastructure details, and producing data insights for research and policy. Part-time (10–30 hrs/week), flexible schedule with occasional Pacific Time meetings.

Analyze and critique new AI benchmarks every two weeks, evaluate methodologies and implications for capabilities, publish and maintain public-facing reports, and examine benchmark tasks and data in detail (using coding agents to assist while retaining independent judgment).