Selected work.

The problem

Purchase orders came in by email as PDFs, and every one was typed into the accounting system by hand.

What we did

A pipeline that reads each PDF, matches it to the right customer and ship-to address, and builds the sales order for the accounting system, where a person reviews and releases it. AI reads the document. Plain code does the matching and scores its own confidence, and anything uncertain is flagged for a person instead of guessed.

What changed

Running in production. About 90% less manual entry on standard formats, and under 30 seconds per document instead of several minutes, checked against real orders by the person who used to type them.

The problem

A support case can come with years of account history in the CRM, and the person answering needs the short version.

What we did

A button on the case that pulls the account’s full history into a panel. The tables of dates and counts come straight from the CRM. AI writes only the short summary on top.

What changed

The whole story on one screen, one click from the case. Approved by CellGate’s support leadership after a live demo, and now in daily use.

The problem

The same customer lived in the CRM, the accounting system, and the customer portal, and the three records didn’t always agree.

What we did

A workflow that compares every account across the three systems and flags only the mismatches, each with a reason, for a person to review.

What changed

One run sorts every account into matched, missing from one system, or in conflict, so cleanup starts from a list instead of a hunt.

The problem

Customer information lived in four places: the CRM, accounting, the customer portal, and a cloud data warehouse.

What we did

The layer of workflows that sits between them. Each one reacts to a change in one system and passes it on where it’s needed, with a log of every run and a hold wherever a person should approve.

What changed

The company’s automations live in one place, in its own account, where its team can open any workflow and read what ran and when.

The problem

How to run AI on company data without that data leaving the building: what hardware to buy, what it could really run, and how to keep the data controlled.

What we did

The plan. We compared the hardware options and recommended one, sized for the models and automations it would host, laid out how to set up a first development machine on tested software, and specified one controlled, logged path between company data and any model.

What changed

A plan to buy and build against: what the hardware will and won’t run, how to set it up, and which uses come first.

The problem

Event Travel International takes the Moroccan developer Chaabi Lil Iskane to buyers abroad. Each expo needs registration, marketing, and logistics to work as one, and they ran on three different systems.

What we did

The public website, online registration and ticketing, the lead records, and the back-office tools the team uses on the floor, plus the ad campaigns and a promo video. Behind it, an offline pipeline that turns booking confirmations and invoices into client itineraries, with every total computed by code.

What changed

Three international expos, more than 1,500 attendees, about 1,500 tickets issued, and no downtime during registration or the events. Still running: the fourth edition opens in Laval, Québec, in October 2026.

The problem

A delegation of Moroccan technology executives is going to CES 2027 in Las Vegas. It needed a public home and a way to keep track of everyone asking to join.

What we did

The delegation’s website, with an interest form in place of an application, and behind it an automation, ready to switch on once enough members have joined, that logs each inquiry to one shared list within minutes. We also help coordinate the delegation’s outreach and logistics.

What changed

A public home for the delegation, and one list of everyone who has asked to join.

The problem

About 50 group-travel events a year, with attendee lists, bookings, and billing coming from separate exports and spreadsheets that didn’t talk to each other.

What we did

The design for one system holding every event, traveler, booking, and bill. Each person sees only what their role needs: agents see their assigned travelers, and fares stay with managers. A traveler gets 24 hours to confirm an itinerary before it expires. After each event, the final bill is compared with what was quoted.

What changed

A complete blueprint, ready to build.

The company’s existing website rebuilt to be easier to use and to update, keeping its own imagery and identity.