When to use/when not to use fast ai

I’ve been playing with the basic Kaggle house pricing dataset recently and was thinking about filling missing values (as spoiler, different missing value columns in the train/test set) and then encoding them in some way (one hot, label…) and was thinking when to use fast ai.

I understand that EDA and some prepping of the data may need to be done outside of fast ai, but are there moments when fast ai isn’t appropriate? My intial thought is would fast ai during categorify detect rare labels?

Any input?