VisionQC is machine-vision quality-inspection software for manufacturing. It runs on a compact edge device at the production line - a Jetson-class GPU module connected to an industrial camera - which does all the image capture and AI inference locally. It inspects every part as it's made and signals the machine to pass or reject it, catching missing components, wrong parts, and unreadable identification codes before they reach the customer. Everything runs on-premise; no cloud dependency, no internet required. Auto-ancillary manufacturers in India — the Tier-2 and Tier-1 suppliers who make components for vehicle OEMs. They face heavy quality penalties from their customers but mostly still inspect by eye, because existing vision systems cost too much, take weeks to deploy, and need a specialist every time anything changes. Today a factory buys an inspection project: an integrator configures a system, it works for a while, it degrades, and re-engineering it means paying the vendor to return. We sell a platform their own quality engineer operates - deployed in days instead of weeks, retrained in-house in an hour instead of a three-week vendor callout, with a complete traceability record that stands up in a customer audit. Value is measurable in their terms: fewer defects shipped, fewer good parts scrapped, inspection labour redeployed. Revenue comes through in two streams. A one-time setup and hardware fee at deployment, covering the edge device, camera, and on-site commissioning. Then a monthly software subscription, metered per inspection point - one inspection decision at one station - rather than per camera or per inference. The recurring fee funds what keeps inspection working rather than merely running: continuous drift monitoring, in-house retraining as their parts and processes change, updated audit and traceability evidence, and support with a 30-minute recovery commitment. A customer who lapses keeps inspecting safely - we never stop their line - but stops improving, as their model goes stale while their factory keeps changing. A supplier starts with one or two inspection points and expands as it works, so revenue grows within an account without a new sale.
80% of factory AI vision projects fail — not because models are bad, but because nothing sustains them. VisionQC is inspection software auto-parts makers own themselves: deploy in days, retrain in-house in hours.