ML + Embedded
VisionQC Inspector
Industrial computer-vision system that flags defective parts on a production line in under 100 ms on a Raspberry Pi.
PyTorch 2.4ONNX RuntimeFastAPIDockerPrometheusRaspberry Pi 5
VisionQC swaps a manual QC station for a Raspberry Pi 5 with a global-shutter camera and a quantized PyTorch model exported to ONNX Runtime.
We optimized the inference pipeline end-to-end — from JPEG decode through bounding-box NMS — to hold sub-100 ms latency at the conveyor's top speed.
A FastAPI service streams results into Prometheus dashboards so plant managers can spot defect-rate spikes the moment they start trending.