Home / Case Studies / Electric Vehicles
Electric Vehicles EV Battery Systems Manufacturer Irvine, CA

35% Scrap Rate Reduction with Computer Vision QA

Custom-trained models catch defects in 200ms that human inspectors miss after hour two

Three product recalls in 18 months. The VP of Quality was on thin ice. Human inspectors were catching defects — but not fast enough, and not consistently enough. The board wanted answers, not excuses about 'inspector fatigue.'

The Challenge

This EV battery enclosure manufacturer was losing $2.3M annually to scrap and rework. Human inspectors performed well for the first 90 minutes of each shift, but accuracy dropped 40% by hour three. With production running 24/7 to meet OEM delivery commitments, they couldn't just 'hire more inspectors' — they'd already tried, and the labor market made it nearly impossible to find qualified candidates.

What We Built

We deployed custom-trained YOLOv8 models at 6 critical inspection points along the production line. High-res cameras capture every component, and edge computing units process images in under 200ms. The models were trained exclusively on this manufacturer's components — not generic datasets. Our overnight ML team retrains models daily on new defect data, so the system gets smarter every single night.

Results

35% reduction in scrap rate (saving $800K in year one)
Defects caught 4x earlier in the assembly process
Inspector accuracy stabilized at 98.2% (vs. 71% human average)
Zero recalls since deployment
2 inspectors reallocated to root cause analysis
Our board asked why we didn't do this two years ago. Honestly? We didn't think AI could handle our specific tolerances. It handles them better than our best inspector on their best day.
— VP of Quality Engineering
12 weeks to full deployment
3 ML engineers + 2 overnight model ops

Facing a similar challenge?

Let's talk about what AI + a supplemental engineering team can do for your business.

Talk to a Dev Lead →