About

We build the systems, and we run them.

Melarex is an AI engineering studio led by Arman. We build production systems end to end: the design, the data, the models, the deployment, and the part most people skip, keeping them running once they are live. The person who writes the code is the person you reach when something needs attention. No agency layers, no account manager in the middle.

The Fomel story

Before we took on outside builds we built Fomel, and it is still the clearest picture of how we work. It reads a company's inbox, its documents, and its numbers, and every morning it tells the CEO what actually needs their attention. We designed the retrieval, wrote the pipelines, and we run the infrastructure it sits on. It is live in production for real companies today. You can look it up.

How we work

Our way of working is probably boring if you are used to agencies, and that is the point. We quote a fixed price, so you know the number before you commit. We care more about a system that runs quietly than a demo that impresses in a room. Every automation keeps a human sign-off until it has earned trust over several clean runs, and only then do we loosen it. When a build is done, you own it. The code, the data, and the workflows are yours, and nothing we hand over depends on us to keep breathing.

The stack we reach for

We are not loyal to tools, only to the ones that hold up in production. Most builds run on Claude for language, Postgres with pgvector for memory and retrieval, and n8n to wire the steps together. Calls and messages go through Twilio, payments through Stripe, and the output lands back in the email, Slack, or CRM your team already opens. For data that cannot leave the building, we build the whole thing to run locally instead.

Where we are

We are based in Europe and work with businesses across the US and the EU. Most of the work happens over a shared screen and a short weekly call. We keep the client list short on purpose, so each build gets the attention it needs.