The mathematical "engineer" (I prefer that to "father") of GR was Bernhard Riemann. And the mathematical "engineer" of QM is David Hilbert. By that I mean the people who introduced the "mathematical scaffolding" that later accomodated the physical theory.
But I don't think either one of them would have come up with the respective physical theories without experimental or theoretical physics input. In fact, when the essential ideas of both theories were formulated, the physicists that did it couldn't imagine the mathematical tools were there already. That realisation, as always, came later.
I think there's always a cycle that goes something like --example: electromagnetism--,
1) Induction: Observation of patterns, or "crude" observation: Lenz, Biot-Savart, etc.
2) Inference of a mathematical or pre-mathematical simple relations: Faraday.
3) The big picture in mathematical terms: Maxwell
4) Experimental confirmation of further predictions: Hertz
Something like that. The history of the development of electromagnetism is a great example of how this works. But, of course, it's more complicated than just that. The different "branches" feed each other in a complicated way.
Once we get the mathematically-closed form of the laws, the great generalisation, it's a matter of pushing and pushing the mathematical model until we find where it contradicts the experiments. It's also a matter of doing more and more refined experiments to check everything's OK.
In the case of quantum mechanics:
1) Wien, Stefan, the spectroscopists (Lymann...), etc.
2) Planck, Bohr, Einstein
3) Heisenberg, Schrödinger, Dirac, etc. find out about a previously-existing mathematical scaffolding -> matrix algebra, Hilbert spaces, Poisson's formulation of mechanics...
4) Anderson finding positrons, which is a prediction of the relativistic version...
Etc.
Sometimes it goes the other way. We find a puzzling experimental discovery, and the theorists must rack their brains, within the mathematical scheme we already trust, in order to understand the unexpected result. If it doesn't, the mathematical scheme must be generalised minimally, ie, in such a way that the treasure of previous results is preserved. Example: discovery of the neutrino. So it's complicated.
We may differ a little bit in what stage is what, but I think we agree in general terms.