Machine learning : Pycaret transformers

Apart of those calling web API, dedicated machine learning transformers are currently missing in FME. 
Pytorch is heavy and requiring gpu while sklearn is also a bit hard to directly set up as transformers (due to the complexity/flexibility of the pipelines). 
Why not try to leverage the work done on PyCaret? Those standardized steps, while limited compared to a dedicated sklearn pipeline, would streamline simple machine learning operations in FME. 
If others have ideas on this, don’t hesitate. I think that those are missing blocks in current FME tools.