Morocco is having an industrial-AI moment. Between the Digital Morocco 2030 push, the automotive and aerospace plants around Kenitra and Tangier, the manufacturing and processing sites inland, and the ports moving more containers every year, there is real money and real appetite for automation. Cameras are already everywhere on these sites. What's missing is systems that actually do something with the footage.
That's the gap I work in. I build computer-vision systems — vision par ordinateur — that run in production. I built and deployed them for heavy industry as an engineer at an industrial AI lab: safety monitoring, measurement, and inspection on live camera feeds. So this isn't a pitch that vision is magic. It's an honest map of where it pays off for Moroccan industry, and where it doesn't.
Where computer vision earns its keep here
Safety monitoring on cameras you already have. Most industrial sites in Morocco already run CCTV or IP cameras for security. The same RTSP feeds can drive PPE detection (is the worker wearing a helmet and hi-vis vest?), restricted-zone alerts, and people/vehicle tracking — without installing anything new on the plant floor. For sites where a single missed incident is expensive, this is often the fastest payback in the building.
Measurement and grading without stopping the line. A camera looking down at a line can size and grade material — parts, produce, packaging — continuously, where manual sampling only checks a handful of times a shift. Across manufacturing and processing, turning a camera into a measurement instrument is a genuinely high-value move.
Quality and anomaly inspection. Defect detection on a line, or "this doesn't look like it should" anomaly flags on equipment, let a small team watch far more than they could by eye. This is where a lot of the automotive and manufacturing interest is heading.
The common thread: vision wins when the information you need only exists as pixels, when the environment is too variable for a fixed sensor, and when a real-time answer is genuinely useful. That's a large space — it's just smaller than the marketing implies, which brings me to the honest part.
Where it doesn't — and I'll tell you so
If a load cell, a temperature probe, or an RFID tag already measures the thing you care about, that sensor will be cheaper and more accurate than inferring the same number from a camera. If one undetected event is catastrophic and there's no backup, vision belongs beside an interlock, not instead of one. And if the hazard you want to catch happens twice a year, you may not have enough footage to train or even validate a model. A first call with me is sometimes me talking you out of a vision project — I wrote more about that in when computer vision is the wrong tool.
"It worked in the demo" is not the same as "it works every shift"
This is the part that sinks projects everywhere, Morocco included. A model that lights up bounding boxes on a clean clip in a meeting room has proven the technology exists — something you already knew. What matters is whether it holds up on your cameras, in your light, at 3am, in the dust and glare of a real site. Building for that — hot-swappable models, real RTSP ingestion, monitoring for drift — is the whole job. It's why I care more about what a proof of concept should actually prove than about benchmark numbers, and why I assume every model will quietly get worse over time unless it's built to be corrected.
The multilingual, on-the-ground advantage
Working in Morocco means the operators, the signage, and the paperwork are in Arabic and French as often as English. Systems that respect that — and a builder who can talk to the plant team in their language — ship faster and get trusted sooner. Being based here, not flying in for a workshop, is part of why the work survives contact with the real floor.
If you're running cameras in Morocco
If you have industrial camera feeds and a problem worth solving — safety, measurement, inspection, or a real-time video pipeline you can't quite get over the line — that's exactly the kind of thing worth scoping together. I'll tell you straight whether vision is the right tool, and what "production-grade" actually takes. Book a call.