SF startup building infrastructure to observe AI agents in production
Full Stack Engineer
You'll spend mornings tracing a flaky evaluation pipeline—Python, Node, a dozen microservices talking past each other—and afternoons rebuilding the dashboard that surfaces those failures to non-technical users. One week it's optimizing PostgreSQL queries against millions of agent traces; the next it's prototyping a new React component for comparing model versions side-by-side. The codebase is young enough that your commits shape patterns, not just patch them.
What they're looking for
- 2+ years shipping production code across the stack, with shipped TypeScript/React and Python projects you can discuss
- Comfort working inside Next.js app router, server components, and the cache boundaries they impose
- Experience designing schemas and writing performant SQL in PostgreSQL, not just ORM abstraction layers
- Track record of debugging distributed systems where the failure is three hops from the symptom
- Genuine curiosity about LLM behavior—not prompt engineering prowess, but rigorous thinking about how to measure and constrain agents