Course, pre-launch

Become the owner of your sovereign AI

The Dompteur learns to handle the tool. The Chaman builds their own.

You leave with your own BrainDup, deployed on a real case from your organisation, and the technical autonomy to keep it running without a cloud provider.

Join the waiting list

Two steps on the same path

These are not two competing products. Chaman d'IA is the upper step of the AS3P training ladder.

Level 1

Dompteur d'IA

Learning to use generative AI day to day: prompting, custom assistants, automation.

See the TNT courses →

Level 2

Chaman d'IA

Learning to design, deploy and evolve your own sovereign AI, with BrainDup and the BrainMad method.

You are here

Who is it for?

  • •Directors and IT leads in small and mid-sized companies.
  • •Consultants and independents in the digital sector.
  • •Technical profiles who want to bring a sovereign AI in-house.
  • •Former Dompteur d'IA participants who want to go further.

Prerequisites

Dompteur d'IA level, or proven ease with generative AI.

No machine learning expertise is required. BrainDup fills a database, it does not train a model.

The format

A hybrid cohort

Four to six weeks, with group sessions remotely and on site.

Your project as the thread

Each participant builds a working BrainDup on a concrete case: their own company's, or one we provide.

Community access

Follow-up and mutual help within the BrainDup community, after the course.

The programme

  1. 1

    The architecture of a sovereign RAG

    The three databases (PostgreSQL, Milvus, Neo4j) and the six open pillars: Django, n8n, FastAPI, Ollama, Docling, GLM OCR.

  2. 2

    The BrainMad method

    The AI-assisted development cycle, from gathering material to the specification, then the epics and the stories.

  3. 3

    Ingestion and governance

    Ingesting documents, managing roles (reader, contributor, curator, administrator), an immutable audit trail, GDPR and AI Act compliance.

  4. 4

    Quality and validation

    The golden dataset (30 cases to learn from, 5 to validate, 15 held back by the client), contradiction detection, entity merging.

  5. 5

    Local deployment

    Going live on a local appliance (NVIDIA GB10 chip and successors), scaling through Docker containers.

What you leave with

✓

A working BrainDup, deployed on your own case.

✓

The code licence (Apache 2) and your contribution to the BrainDup project.

✓

Access to the community and to updates of the method.

The only course that makes the participant the owner and independent operator of their company AI: no token costs, no sensitive data exposed, no dependency on a cloud giant.

Pre-launch

Join the waiting list

The first cohort starts in autumn 2026. Those on the waiting list get founder pricing and priority access.

Pricing is by quote: every project is scoped in a call, so the support fits one person, a team, or the whole company.

No machine learning expertise is required: BrainDup fills a database, it does not train a model.

Which corpus, which department, which concrete problem would you hand to your BrainDup?

Fields marked with an asterisk are required. Reply within 48 working hours.

Would you rather talk it through? Book a slot for a demonstration of BrainDup and the method.

Project contact: Eric Vandeweyer, BrainDup project lead.