96%defect detection rate in trial
-60%customer-reported defects
< 2%false reject rate
About the client
An auto-components manufacturer, based in Pune, Maharashtra, operating in the manufacturing & msmes sector.
The problem
Manual inspectors checked thousands of machined parts per shift, and fatigue led to defects reaching the customer, triggering chargebacks.
What we did
- Camera and lighting setup on one inspection station
- Model trained on labelled images of good and defective parts
- Edge device running inference at line speed without internet
- Reject signal and dashboard for supervisors
- Shadow-mode comparison with human inspectors before go-live
Outcome
- 96% defect detection rate in trial
- -60% customer-reported defects
- < 2% false reject rate
Services involved
Computer Vision Development · AI Automation
Client identity withheld. This anonymised case study is an illustrative example of our approach; figures are indicative and results vary by business, market and budget.
