Project case studies with role, constraints and results: IoT-DWS integration across 16 sortations and AWB detection with computer vision.
IoT-DWS Integration System
Designed and deployed an integration system between IoT devices and DWS (Dimensioning Weighing Scanning) machines across 16 sortations.
Senior Software Engineer — System Owner
Context
Took over and rapidly mastered an IoT-DWS system after the previous responsible team member left the project. Needed to quickly understand the current architecture and continue developing the integration solution.
Problem
Needed integration between IoT devices (sensors, cameras) and DWS machines — a complex physical system with multiple different protocols (HMI, FTP, HTTP).
Solution
Designed an HTTP-based integration engine with HMAC security, RocketMQ messaging, and protocol adapters for HMI/FTP/HTTP.
Architecture
IoT Devices ↓ Protocol Adapters (HMI/FTP/HTTP) ↓ HTTP Engine + HMAC Security ↓ RocketMQ ↓ DWS Machines
Challenges
Multiple different protocols, security requirements (HMAC signature), real-time data processing, production reliability.
Result
Solution deployed across 16 sortations, stable in production, supporting 3 protocols (HMI, FTP, HTTP).
AWB Detection / Computer Vision
Applied Computer Vision and Machine Learning to automatically detect and OCR shipping labels from camera images on sortation lines.
Senior Software Engineer — ML Engineering
Sortation machines capture images of packages as they move on conveyor belts. Need to automatically read AWB (Air Waybill) from images for tracking and sorting.
White label on white package — traditional CV methods are not robust enough. Needed ML-based detection.
Pipeline
Image ↓ AWB/Label Detection (YOLO) ↓ Bounding Box / OBB ↓ Rotation Correction ↓ Crop ↓ OCR ↓ Text Extraction
White label on white package, rotated labels, varying lighting conditions, real-time inference requirements.
Model fine-tuned for AWB detection, processing 70,000 images/month. Presented at Tech Talk workshop.
Dong Nguyen