A promising framework for flexible, simple MLOps.

A friend pointed me to this the other day. ZenML is part of a growing mass of MLOps tools designed to make the building, training, deploying and maintaining of ML models easier.

The framework provides some decorators to convert your normal python code into a pipeline, ZenML then makes it (fairly) easy to test those pipelines locally and deploy onto cloud infrastructure. In theory, this sounds great.

Without doing a full review of the framework, my main concerns are: do I want to tie my ML deployment to a young open source project? Does their model of ML deployment fit with my specific use case? How much new stuff do I have to learn to be able to even test it out?

I guess these are challenges to any new entrant, but does anybody have good answers to these questions?