A Canadian industrial AI company that builds the seeing part of inspection machines.
Pindeed Inc. develops the software, detection algorithms and models behind AI fabric inspection, and supplies them to the companies that build the machines. Our engineering runs between Canada and China; our deployments run in Chinese mills.
One region in a line-scan field, and what the system reads from it. The size is a real measurement from the deployed line.
Machine builders reach the mills. We make the machines see.
Textile mills buy inspection machines from the people who have built them for decades, who service them, and who know the cutters, the tension, the rollers and the dealers. That relationship is not something a software company should try to replace.
What those builders cannot easily hire is a team that keeps a detection model right on a real line for months: through lighting drift, saturated backings, new fabrics every day, and operators who need the system to learn from their calls. That is what we have spent our engineering on, and what our field records show.
So we supply the intelligence and the integration, and the machine builder supplies the machine. We measure ourselves by what the machines find on the floor, not by demo footage.
Pindeed is Pin + Deed.
Pin: pinpoint the problem
The engine finds the fault and pins it down: what it is, how large, and where on the roll. Size and position are both to the millimetre, with the position taken from the machine's own encoder.
Deed: turn insight into action
A finding is worth something only when the line acts on it. Detections go straight into the work on the floor: the operator confirms on the touch screen, cut points and grading follow, and the roll records and reports carry them downstream. Inspection that runs in production, not analysis that stops at a chart.
A promise matters only when it becomes a deed.
Precision. Integrity. Deeds.
Measured on the real line
Type, size and position to the millimetre, read from production frames rather than lab samples. A finding that cannot be located cannot be acted on.
Evidence over demos
Every claim on this site traces to a database on a running machine. When we cannot show a number, we say so.
Built for the 3 a.m. shift
Imaging health monitors, over-the-air updates and a machine that keeps inspecting when the network is gone. We would rather ship one machine builder's product line properly than sign ten letters of intent.
The engine is built for roll materials. Textile is where it is proven.
The same acquisition, normal-model learning and band detection apply to non-wovens, paper, film and foil. We take pilots in those materials with equipment partners who bring a line; we do not claim them until a real line has run.
We are hiring.
Engineers who care about the quality of industrial software and want to see their work on a factory floor.
Computer vision · Industrial Go / Python · Machine integration · Field application engineering.
careers