We partner closely with frontier AI and deep tech teams, working as an extension of their hiring process and often seeing roles before they’re public.
The roles below are a snapshot of what’s live. If nothing fits today, upload your CV to stay in touch as new opportunities come up.
Design and own the system that finds exceptional founders before anyone else. Ranking models, entity resolution and a knowledge graph built over the open web and 15 years of a top early-stage fund's data.
Own the evals framework a frontier AI company uses to decide what ships. ~50 benchmarks today, 100–200 soon, across web, desktop and the command line.
Build zero-to-one products in a small pod at a late-stage AI company moving from voice agents to an AI-native platform for US housing. Full-stack, on-site, growing above 100% a year.
Own the data layer at a fast-growing AI company, from pipelines and quality checks to the core models the whole business builds on. Small New York team, high ownership, on-site.
Train the models behind computer-use agents with SFT, online RL and offline RL across GUI, API, MCP and coding domains. The hard part is rewards that hold up on long-horizon tasks.
Build the environments, tasks and reward signals that agents are trained on. Engineering-heavy work at a frontier AI company, with hundreds of apps and tens of thousands of tasks to extend.