My customer only need MLOps, so I try to plan a hardware architecture and resource capacity for MLOps on-premise. Does any one have some recommendation for me?