Sounds great. Something like a tidyverse version of the data pipeline capabilities of scikit-learn would be awesome.