As a Computer Vision Engineer, I developed an end-to-end monitoring system using YOLOv7 and OpenCV for smart city infrastructure. I optimized the pipeline for resource-constrained edge devices, achieving a high-performance throughput of 25 FPS at 85% accuracy. To ensure production reliability, I implemented custom NMS (Non-Maximum Suppression) and used MLflow for rigorous experiment tracking and model versioning. The final system provides scalable, low-latency vehicle detection and license plate recognition for real-time traffic analysis.