Acquisition
Getting data where there is no API. 45,000 files, 56 GB.
Independent AI product engineer — Casablanca · on Paris time
Whole AI products — from the scraper to the invoice.
Everyone can build the demo. What stops it is the rest: multi-tenancy, permissions, quotas, cost per request, audit trails, job queues, evaluation. I built all of that alone, and it runs in production.
Projects from €12,000 · Day rate €600 · available now
Demo — tender file no. 42/2025 · 312 pages · FR/AR scanned
Query
What is the bid bond, and by what date?
Answer
The bid bond is set at 150,000.00 MAD, to be lodged no later than 14 November 2025, 10:00. The required certification is sector E — class 3.
3 values · 3 sources · 0 unsourced
Sources — click a citation
A reconstructed demo built from a typical file, values anonymised — this is not a client document. In production: 274,376 citations across 58 fields, 99.95% of them carrying the verbatim passage.
Counted in the code — BidTender, in production since Sept. 2025
84API routers in production
300business-logic modules
274,376provenance citations
1person, from the scraper to the invoice
Why an AI demo doesn’t become a product
Three reasons, and none of them is the model.
- 01
It only has one user
A demo runs for you. A product runs for customers who must never see each other. Isolation belongs in the database, inside the query itself — the naive implementation that searches everything and then filters is secure, and broken: recall collapses as the corpus grows, and the bug never shows up in a single-tenant demo.
- 02
Nobody knows what it costs
While it’s a demo, the model bill is a rounding error. In production it’s your margin. You need a known cost per request and per customer, quotas, ceilings, an overrun alert — and routing that sends simple questions to a far cheaper tier without anyone noticing.
- 03
Nobody knows whether it’s right
“Is this better?” judged by eye on a handful of familiar queries is not an answer. You need an annotated set that belongs to you, verification of answers against their sources, and the right to say “that information is not in this file” — because the only truly serious failure mode is a system that always answers.
The whole chain
— what “from the scraper to the invoice” actually meansUnderstanding
Bilingual OCR routed page by page. 1.1 M located words.
Decision
58 extracted fields, scoring, inference — source alongside.
Delivery
Interface, exports, generated documents and decks.
Permissions
Per-tenant isolation in the database, (feature, role) matrix.
Billing
Plans, usage counters, quotas, invoices.
Three offers, not a catalogue
The three offers in detail →Each answers a specific moment. If your situation fits none of the three, I’ll say so on the first call.
- 01€14,000 – 22,000 · 4 to 6 weeks
From demo to production
Your POC works on a laptop and nobody dares put it in front of customers. I add what’s missing to make it sellable: isolation, permissions, quotas, controlled cost, evaluation, guardrails, observability.
- 02€30,000 – 55,000 · 2 to 4 months
The whole product
You want an AI product and you have no team. Acquisition, processing, database, AI, interface, permissions, billing, production. One person accountable for the whole chain.
- 03€9,000 – 15,000 · 3 to 4 weeks
The document foundation
Your documents are heavy, scanned, multilingual, contractual, and the AI answers beside the point or without proof. Ingestion, OCR, hybrid retrieval, verifiable citations, an evaluation set that belongs to you.
Four references
See all →BidTender — multi-tenant SaaS
A whole AI product, built and run alone
A multi-tenant SaaS for bidding on Moroccan public tenders: the daily scraping that fetches the files, the bilingual OCR that reads them, citable retrieval, structured extraction, decision engines, generation of the documents to submit, the sales funnel and billing. 84 API routers and 300 business-logic modules, written and operated by one person.
1 person84 API routers · 300 business modules · in production
Leyton — consulting engagement
Real-time personalisation with a RAG persona pipeline
A cosmetics platform was recommending from frozen segments while ignoring all the text it already owned. Moved to per-person, real-time recommendation built on retrieval — and explainable by the business team.
+15%conversion rate · −20% analysis cycle
Chemin du Roi — industrial platform
Digital twin of an anaerobic digestion plant
Five heterogeneous real-time streams, five protocols, five notions of time — unified into one platform that plant engineers make decisions on every morning.
5 → 1protocols unified · OPC UA · Modbus · MQTT · SFTP · API
Lbassi · Wadrari
Two mobile products designed, built and shipped solo
What these two apps prove comes down to one word: I finish. A client handing over a document system fears receiving an impressive notebook and nothing that runs.
Shippednot a prototype · Kotlin · Expo · Supabase RLS
The demo works. Now what?
Describe the situation in three lines — what already exists, what has to exist, and by when. In thirty minutes I’ll tell you which of the three offers it falls into, or that it isn’t for me. Reply within 24 hours on weekdays.