The complete checklist we use in our ML architecture reviews - training infrastructure, data pipelines, model serving, …