Ilyas Ouhnine AI products · end to end

About

AI engineer, Centrale-trained, based in Casablanca. I build whole AI products — from the data you have to go and fetch, to the invoice you send the customer.

I’ve spent the last two years on two systems that had to run for real: a multi-tenant SaaS for public-procurement responses, which I co-founded, and a digital-twin platform for an industrial plant. In both cases the interesting part was never the model. It’s what happens when the input data is ugly, when a stream drops, when a document runs four hundred pages — and when you still have to invoice somebody at the end of the month.

On the first one, I wrote and I operate 84 API routers and 300 business-logic modules: the daily scraping that fetches the data, bilingual OCR, citable retrieval, structured extraction, decision engines, document generation, per-tenant isolation in the database, cost-based model routing, audit and billing. Alone. That is what sets me apart: not a technique, a scope.

Five projects, five times the same move: a source nobody wants to touch — a site with no API, five industrial protocols, scanned PDFs in two languages — turned into a product that runs, with a database, permissions and an interface. That’s useful to teams too small to hire three specialists, and to teams who already have a prototype that won’t scale.

I work in scoped engagements, remote, in French or English. Scoping is done by deliverable and delivery date rather than days per week — which is what gets you a firm date instead of a theoretical availability.

At a glance

Education
École Centrale Casablanca — Data Science & Digitalisation · exchange at CentraleSupélec
Languages
French (DALF C1) · English (TOEIC 885) · Arabic (native)
Based in
Casablanca — same time zone as Paris
Availability
Immediate — remote engagements
Invoicing
Moroccan sole-trader status, invoices in euros

What I don’t do

No public procurement or tendering — that’s my own company’s field. No digital twins or industrial AI — that’s my employer’s. And no project where nobody can tell me how we’ll know it worked.