Technology
Edge AI Platform Architecture
From capture to quality intelligence — a modular, edge-first platform designed for factory floor deployment.
Capture Layer
Phone cameraUSB cameraIP cameraEdge device camera
Inference Layer
ONNX runtimeYOLO modelsMulti-model routerConfidence scoring
Classification Layer
Defect taxonomySeverity scoringRoot cause analysisQuality gates
Decision Layer
Operator confirmationHuman overrideAlert dispatchPass/rework/scrap
Storage Layer
Local SQLiteSHA-256 audit chainImage evidence storeConfigurable retention
Sync Layer
Encrypted cloud syncOffline-first queueBandwidth managementDelta sync
Dashboard Layer
Real-time metricsDefect trendsSupplier scorecardsAlert feed
Key Capabilities
Multi-Model Routing
Route inspection images to the right detection model based on product category and defect taxonomy.
Severity Scoring
Classify detected defects by severity — critical, medium, or low — for prioritized response.
Audit Trail
Every inspection decision records an immutable event with SHA-256 hash chain verification.
Real-Time Alerts
Threshold-based alerts for defect clusters, shift trends, and quality metric breaches.
Human-in-Loop
Operators confirm or override detections. ML never makes autonomous release decisions without approval.
Pilot Reports
Automated pilot report generation with detection metrics, operator feedback, and quality improvement analysis.
Supported Edge Hardware
Raspberry Pi 4/5
Entry-level edge inference
Jetson Nano
Production edge deployment
Jetson Orin
High-performance edge
USB Camera
Common USB camera support
IP Camera
RTSP/ONVIF camera support
Phone Camera
Mobile app capture