ML Observability

When is the best time to start thinking about ML observability in your opinion? Yesterday, like in terms of the model development lifecycle, I would say that you have to have observability when you deploy a machine learning model. In most cases, when you hear MLOps observability, you hear everything post deployments. The top priorities for an observability system are that they need to give you continuous monitoring. It needs to be complete. So it needs to give you a view of the entire system.