Description Develop and optimize GPU accelerated video inference pipelines, including batching, stride control, and throughput tuning. Implement, evaluate, and improve object detection models (YOLO or similar) and build temporal smoothing/tracking logic for safety event detection. Optimize model performance using TensorRT, ONNX, CUDA, and GPU profiling tools to maximize throughput and minimize latency/VRAM usage. Build and maintain integrations with event-driven APIs, Azure Event Hub, Blob Stora