End-to-end MLOps architecture built for polyp segmentation — featuring distributed Ray training, MLflow experiment tracking, and automated CI/CD with Kubeflow Pipelines and KServe (Triton) deployment on Google Kubernetes Engine.