Founder and engineer. Questions come straight to me.

Who built this

I am Oleksii Liashuk. I write the detection code, I answer the email, and I run the integration calls. There is no support queue to be routed into, because there is nobody to route you to.

Before Teuscan I built a production computer-vision system for vehicle damage tracking — the same problem shape as freight, on a different object. That is the background this is built on, and it is why the answer to "will it work on my data?" is an upload rather than a slide.

Why a vision model instead of fixed cameras

The established approach in this category is hardware: mount cameras over the dock door, control the lighting, control the angle, and run a detector trained on what those cameras see. It works, and it costs a quarter of procurement plus an installation before anyone finds out whether it helps.

Freight damage is the wrong problem for that. A crushed corner, a torn wrap, a leaning pallet, a punctured carton, a shifted load — the failure modes run into the thousands of forms, and a classical detector only knows the ones it was trained on. A vision model reads the photo the way your receiving clerk does, which means it generalises to damage nobody thought to label. That is the whole technical bet, and it is why Teuscan needs no hardware: it works on the photos your team already takes.

Where this came from

Warehouse managers told me about it, more or less unprompted. Photo documentation for claims defense has been standard since around 2018 and it is a solved problem — that is not what this is. What is not solved is that a person still has to notice the damage in the first place, on a busy shift, and get a usable photo of it. One manager had gone as far as collecting hundreds of images each of damaged boxes, damaged pallets and damaged wrap. People do not do that unless the problem is real.