A real-time AI vision pipeline built with YOLOv8 and NVIDIA TensorRT for GPU-accelerated inference. This project performs live object detection and applies multi-layer, privacy-preserving face anonymization in video streams. Designed with performance, security, and deployment readiness in mind.