I developed an automated pipeline for training, deploying, and monitoring Ultralytics YOLOv8 models to optimize video annotation workflows. This project included automating CVAT hosting with Ansible, improving model scalability on Triton servers, and enhancing inference speed through multithreading and multiprocessing techniques. By streamlining model deployment and automating key processes, the solution significantly reduced manual effort and boosted Mouver's ability to efficiently analyze large-scale video data.

I researched and implemented the ECON model for 3D virtual try-on, containerized it with Docker, and deployed it on an AWS EKS Kubernetes cluster with optimized settings and autoscaling for efficiency.