Explain a model deployment architecture you would use to support production machine learning workloads.
