Optimize AI/ML workloads for sustainability: Part 3, deployment and monitoring