mlops1.md

Challenge	Automation solution
Training-Serving Skew, where differences crop up between training and serving stacks	Feature Store
Clients are interested in a capability that involves multiple steps, only one of which is the ML model	ML Pipelines
The incoming data changes characteristics	Continuous evaluation
The correct answer changes over time	Continuous training
Model fairness issues arise	Mumbo Jumbo
Parts of the system aren’t changed when new code is checked in	CI/CD
Need to be able to troubleshoot old results	Model registry, Data registry