Firecrawl is the web data API to search, scrape, and interact with the web at scale. Turn any source into clean Markdown or structured data your agents can ship with. Backed by Y Combinator.
About the role
Own the internal AI stack: the agent harnesses, configs, shared skills, and model routing every team uses. Set them up, keep them sharp, and ship changes as the tools change Start with the revenue team. Learn how they actually work, find the jobs an agent should be doing, and get them running on it, then go team by team Found on 1752vc Careers, the job board for startup and VC roles.
What they're looking for
- You've rolled out coding agents or AI tooling across a team and can show what changed after. An adoption, not a pilot
- EQ. You can tell the difference between someone who's resisting and someone who's scared, and you know that most "I don't have time to learn this" is the second one. You bring people along without making them feel behind
- An engineer or a very technical operator who lives in Claude Code, Cursor, or Codex and writes code. When the harness is missing something, you build it
- You think about cost per useful token, and you've looked at spend data and changed something because of it
- You teach by showing. Pairing for an hour beats writing a guide nobody reads
- AI-pilled, bias to action, all in. You believe this is how work gets done now and you're already operating that way
More about this role
Firecrawl is looking for someone to own how we use AI internally. The thesis behind the role: a company's output used to scale with headcount, and ours scales with tokens, meaning how much compute we run and how well we turn it into shipped work. A small team that does this well operates like one several times its size, and the ceiling on that isn't budget, it's adoption. You'll set up and maintain the agent harnesses, configs, and model routing every team runs on, revenue first, get the people who barely use agents onto them, and work straight from our token spend data to find who's burning it without output and who isn't using enough.
Most of this work is people, not config. Getting someone to hand real work to an agent means asking them to change how they've worked for years, with tools that feel unfamiliar and sometimes threatening. You'll need EQ as much as technical depth: reading who's stuck and why, meeting people where they are, and staying with them until it sticks. You'll still live in the tools yourself, write the code when the harness falls short, and teach by showing. You'll be measured by who's using agents well now who wasn't before.
Equity Range: Competitive...
Browse similar: AI jobs · AI startup jobs · Startup jobs · Remote jobs · San Francisco Bay Area