An SRE's guide to optimizing ML systems with MLOps pipelines