What you'll do
You'll embed with Anthropic's largest enterprise customers — Fortune 100s, sovereign-AI buyers, regulated industries — and ship working Claude deployments inside their environments. Not pitch decks. Not POCs that stall. Working systems that pass evals against their data and survive contact with their compliance teams.
A typical month looks like: two weeks on-site at a customer, mapping their workflows; one week building skills, MCP servers, and evals against the workflows you mapped; one week shipping it and training their internal teams to own it after you leave.
What we're looking for
- 5+ years shipping production software at companies with real users
- Comfort writing prompt scaffolds, MCP servers, and skill manifests — not just calling APIs
- Eval design experience: you know how to measure if an agent is actually working, not just demoing
- Stakeholder fluency: can translate between a CFO and a Claude API request
- Travel tolerance: ~30% time on customer sites, mostly North America
Bonus
- You've shipped agents into a regulated industry (health, finance, legal, gov)
- Public writing or open-source work in the agent / MCP ecosystem
- Experience running technical workshops for non-technical stakeholders
Six of the eight Operator skills, explicitly required.
This is the canonical FDE pattern — Anthropic isn't hiring you to research models or build SDKs. They're hiring you to make Claude actually work for a specific customer, end to end, and own the outcome.
Compensation & benefits
$200K – $300K base. Equity refresh annual. Anthropic covers 100% medical / dental / vision premiums for employee + dependents. Unlimited PTO with a 4-week minimum-take floor. $5K annual learning & development budget. Relocation assistance to any of the listed offices.
How to apply
Apply directly on Anthropic's Greenhouse board (link above). The hiring manager is reviewing applications weekly. If you're a current subscriber to The Operator and want a warm intro, reply to any newsletter and mention this role — we maintain a notes channel with several Anthropic FDEs who can flag your application internally.