Custom object detection, tracking, and image analysis pipelines using YOLO, RT-DETR, OpenCV and PyTorch. From data annotation and model fine-tuning through to a deployed, running system. I've built a production CV pipeline processing 120K+ frames at 25 FPS.