AI Engineer (Intern)

We're building something hard, in stealth. We need one person who can take a detailed-but-fuzzy spec and build the entire thing end to end — pipeline, data model, integrations, server, and the thin web surface on top — with no PM in between and no one above you deciding what "done" means. This is not a role where you own a slice. You own the whole loop, ship it, deploy it, and get paged when it breaks. Skills you need AI / LLM engineering. You build with LLMs, not around them — parsing messy inputs, inferring schemas, extracting structure, having the model judge and self-check its own output, and thinking in confidence and precision rather than "it looked right." Not just calling a chat endpoint. Integrations. Fluent with official third-party APIs — OAuth flows, webhooks vs. polling, and turning messy real-world events into a clean, normalized store. Backend. Python or Node. Clean data modeling, schema validation, sandbox/test environments, API and server design. Verification instinct. Your default question is "how do I prove this mechanically," not "it looks right to me." You reach for gradeable pass/precision criteria before you trust anything. Deployment. Docker, basic CI/CD. You ship your own code to prod without waiting on infra. Data literacy. Comfortable defining and reading pass/precision rates, accuracy, and drift signals — and using them to decide what ships and what gets killed. Systems judgment. You can hold a whole spec in your head, make the architectural call, resist over-building, and protect the invariants that can't be bolted on later. Bonus: prior 0-to-1 / founding-engineer experience, and hands-on MCP work. What we actually want to see Not a resume of responsibilities. Proof of work. Show us things you personally built and shipped, end to end — repos, live products, side projects, anything real we can look at and poke. We care about what you've actually made, not where you've worked or what you were "part of." One thing you built alone that works beats a decade on teams that shipped around you. The filter This is simple. If you can build this kind of thing entirely by yourself, end to end — apply. If you can't, please don't. We're not looking for someone to grow into it; we're looking for someone who can already do it.