Build AI that actually ships
We're a small, senior team building production AI agents and knowledge systems for enterprise and the public sector: real systems, real users, real stakes. We don't always have formal openings, but we're always interested in people who build.
Ship over demos
You care about software that works in production for real users: correctness and reliability over flashy prototypes.
Comfortable across the stack
From LLM orchestration and retrieval to the API and the UI that exposes it. You go where the problem is.
Hard problems, real constraints
You like regulated, high-stakes domains where data discipline, security and traceability actually matter.
What you'd work on
There are two fronts and nearly everyone touches both. On one side our own products, running today on critical data across insurance, finance, hospitality, healthcare and engineering. On the other, the systems we build inside companies and the public sector, where the software has to live alongside whatever is already there.
From orchestration to the screen
The whole path: how the model is orchestrated, how context is retrieved, what the API exposes and what the person using it ends up seeing. Nobody inherits a disconnected slice.
Systems already in flight
Most of the work lands on live things, with users inside and data that matters. It looks little like starting from scratch: it forces you to read before you write.
Constraints that aren't negotiable
Domains where traceability, data protection and what happens when something fails are part of the brief, not a later phase.
The terms, before you have to ask for them
We'd rather put them in writing than have you drag them out of a call.
Remote
The company is in Madrid and the work doesn't depend on being there.
Small, senior team
Nobody learns the craft on a client's project. It also means there's no middle layer standing between you and the problem.
The code gets handed over
We close projects leaving the code, the documentation and the reasoning behind each decision. Whoever hires us can carry on without us.
Technology is chosen on fit
Not on how recent it is. A good part of the craft is ruling out what was never needed.
How applying works
- There isn't always an open role. A spontaneous application is the normal route, not the exception.
- Tell us what you've built and what your part in it actually was. If there's something we can look at — a repository, a system in production, something you've written — link it.
- The Naktor team reads it directly, with no commercial filter in front.
Introduce yourself
No formal opening needed. Tell us what you've built and why you'd be a fit: you'll reach the Naktor team directly.