help for Kaggle Titanic Set fill the missing Age by median age of Pclass and Sex

[scidam]

You are probably looking for this:

train_df.Age.fillna(train_df.groupby(['Sex','Pclass]).transform('median').Age, inplace=True)

# from now train_df.Age doesn't contain nans

I would suggest you to take into account 'title' property, e.g. Masters are young people, etc.
Another suggestion is to use combined dataset (from train and test ones) to get 'median' estimations, i.e.
something like this

train_df.Age.fillna(pd.concat([train_df, test_df]).groupby(['Sex','Pclass']).transform('median').Age.iloc[:train_df.shape[0]], inplace=True)