Most of today’s ML models are built using public datasets and trained in notebooks. No surprise these models are hard to productize. Operations require automation. If the development loop is not automated, it’s broken. The good news, MLOps solves it. Bad news, there’s a lot of fluff.