Selectively applying pyspark transformers to dataframe rows

Question

I have a pyspark dataframe with the following schema

+-----------+---------+----------+-----------+
|     userID|grouping1| grouping2|   features|
+----------