ID   score
100    1
100    2
102    6
102    8
102    12
103    4
103    7

ID   score  result
100    1      NaN*
100    2      1
102    6      NaN
102    8      2
102    12     4
103    4      NaN
103    7      3

df['result'] = df.groupby('ID')['score'].transform('diff')