Grouped Time-Series with ffill

import pandas as pd
import numpy as np

dates = pd.date_range('2024-01-01', periods=20, freq='D')
df = pd.DataFrame({
    'date':   dates,
    'device': ['A','B'] * 10,
    'value':  [np.nan if i % 3 == 0 else float(i) for i in range(20)],
})

df = df.sort_values(['device','date'])
df['value_filled'] = df.groupby('device')['value'].transform(lambda s: s.ffill())
print(df)