AI consultancy and engineering

AI on your data, on rules you can check.

I help teams find where AI pays off in their own work, write down what may reach a model, then build it. The four rules below come with every project.

The rules

  1. 1

    Personal data is redacted before any model sees it.

    Names, ID numbers and bank accounts become placeholders, restored only on your side.

    Show evidence
    Fig. 1 · accounting toolkitTo , ID no. . Deadline: 15 days.
  2. 2

    We say where processing happens.

    Per project, in writing: EU vendors, your own machine, or our servers in Germany.

    Show evidence
    Fig. 2 · Chelar“I took 5 rolls from the cellar”
    speech → EU · audio discarded after transcription
  3. 3

    A person approves every output.

    The model drafts. Your accountant or engineer sends.

    Show evidence
    Fig. 3 · accounting toolkitDraft note to [Client-1] · waiting for approval
  4. 4

    What it cannot handle, it hands back.

    Unknown cases go to manual review with a reason. Nothing is guessed.

    Show evidence
    Fig. 4 · accounting toolkitScanned notice unreadable → manual review, reason recorded

Where I work

How a project starts

  1. 1 · Call30 minutesYou describe the work that eats your week.
  2. 2 · VisitOn siteI sit with the people who do it and learn their words.
  3. 3 · Rules and proposalOn paper firstWhat may reach a model, where, who approves. A prototype where it helps.
  4. 4 · BuildQuoted per projectBuilt with agents, tested first, reviewed before it ships.

Earlier clients: imec, Knowledgeflow, Schoolmakers.

Tell me what takes your team's week.