Practical notes on how to break a Machine Learning system into smaller services to make it easier to deploy, monitor, and scale.