How to find the last non zero element in every column throughout dataframe?

You can convert 0 to missing values, use forward filling and select last row by indexing, last cast to integer:

df = df.mask(df==0).ffill().iloc[[-1]].astype(int)
print (df)
    A  B
5  10  2

Here's one approach using ndarray.argmax and advanced indexing:

first_max = df.values[df.ne(0).values.argmax(0), range(df.shape[1])]
out = pd.DataFrame([first_max], columns=df.columns)

---

df = pd.DataFrame({'A': [0,0,0,10,0,0] , 'B': [0,2,0,0,0,0]})

first_max = df.values[df.ne(0).values.argmax(0), range(df.shape[1])]
# array([10,  2])
pd.DataFrame([first_max], columns=df.columns)

    A  B
0  10  2

---

Update

In order to find the last nonzero:

row_ix = df.shape[0]-df.ne(0).values[::-1].argmax(0)-1
first_max = df.values[row_ix, range(df.shape[1])]
out = pd.DataFrame([first_max], columns=df.columns)