columntransformer

I push nearly all preprocessing into a Pipeline so training and inference paths share exactly the same logic. ColumnTransformer is the workhorse here because real-world tables mix numeric, categorical, boolean, and text fields. It gives you reproducib