Pandy normalizują grupę
df.groupby('indx').transform(lambda x: (x - x.mean()) / x.std())
Careful Curlew
df.groupby('indx').transform(lambda x: (x - x.mean()) / x.std())