Intelligence within.
Intraion is an AI infrastructure practice. We get a company's data into a state worth querying, then build the systems that query it — by default inside their own infrastructure rather than inside somebody else's cloud.
The practice is built for European firms handling material that cannot leave the building — legal, financial, and professional services, where GDPR and the EU AI Act are not optional extras but the operating conditions.
Everything we build is open source. Every engagement ends with documentation, source code, and full control in your hands. No licences, no subscriptions, and no reason to keep paying us once it works.
Pierre-Yves Dehon
Founder · AI Infrastructure Consultant
Ten years across infrastructure, DevSecOps, and cybersecurity, now spent full-time on getting AI to run safely inside other people's networks. The security background is the reason this practice exists: most AI deployments are architected as if the data were harmless.
A background in teaching shapes how engagements end. The goal is never a system only we understand — it's a system your team can operate, extend, and audit once we're gone.
Based in Portugal. Working remotely across the EU.
Four commitments we don't negotiate.
Open source first
Every tool we recommend and every system we build is open source. You own it outright — no licences, no subscriptions, and no vendor in a position to pull the rug or change the terms.
EU sovereign by default
Your data stays in the EU. Your infrastructure stays on European soil. Sovereignty isn't a feature we bolt on at the end — it's the constraint we design from.
Sovereign by default, not by dogma
For AI touching sensitive material there is no responsible alternative to running it yourself, and that is most of what we build. But a public model over non-sensitive data is sometimes simply the better answer, and pretending otherwise would cost you money. The skill is knowing which is which — and being able to show the reasoning to a regulator.
Built in public
Whatever we run for ourselves, we document and publish — architecture, configuration, and the mistakes. The blog is a working audit trail of how we build. Read it before you talk to us.