Two semiconductor chips are optically aligned to form a hybrid semiconductor device. Both chips have optical waveguides and alignment surface positioned at precisely-defined complementary vertical offsets from optical axes of the corresponding waveguides, so that the waveguides are vertically aligned when one of the chips is placed atop the other with their alignment surface abutting each other. The position of the at least one of the alignment surface in a layer stack of its chip is precisely defined by epitaxy. The chips are bonded at offset bonding pads with the alignment surfaces abutting in the absence of bonding material therebetween.