# By default, it returns the index for the maximum value in each column. df.idxmax()
df.idxmax(axis="columns")