Nelfet.
careers
Build production AI with a small senior team.
We hire rarely and carefully. If you build software that holds up with real users and want to work close to the model, we'd like to hear from you.
culture
How it feels to work here.
Small teams, real ownership
No hand-off layers. Everyone on a project knows the full system and shapes technical direction.
Close to the model
Engineers understand the AI layer: capabilities, failure modes, cost, and how to evaluate it systematically.
Production is the point
Evals before shipping, instrumentation before incidents. The standard doesn't relax after go-live.
Honest by default
Name trade-offs clearly — to clients and to each other. Clarity is more useful than comfort.
what we look for
People, not checklists.
- Engineers who own what they ship, including observability and post-launch behaviour.
- People who evaluate and measure AI — retrieval, context, evals, real failure modes.
- Designers who work directly in the medium and hold performance to the same standard as aesthetics.
- Generalists who move across the stack: problem-shaping through architecture, implementation, and instrumentation.
- Anyone who prefers precise measurement over confident assertion.
how we work
The shape of the work.
- Remote-friendly. Async by default, sync for decisions that require it.
- Small senior teams — lean enough that everyone knows the full system.
- Frequent, reversible releases. Feature flags and gradual rollout are standard, not exceptions.
- Evals and observability built into every AI project from the start.
- Direct, written communication. Trade-offs documented, decisions recorded.
No open role that fits? Tell us anyway.
We don't always have a posting up. If you're the kind of person described here, send a note and a few things you've built.