General questions on some opportunities in CUDA

Hello! I’m new to the CUDA. Actually, except video card and curiosity I do not have anything. But before start to learn CUDA i want to know can CUDA make that i need. I plan to write a program where the GUI will be fully vested in Unity3D and the calculated part will be somewhere else, for example in CUDA.
Moreover, I would like to put two video cards on the computer (not in NVIDIA SLI mode, just two separate cards) so that one was fully engaged under Unity3D and the other is completely under calculation process.

The bottom line is that Unity is very useful for visualizing different objects, because it has a lot of tools to create these objects and manage them, but it is not very good in the massive computing of the behavior of these objects. I do not mean the standard game engine physics, it works fine, but that physics that I came up with is unlikely to be run quickly. So I want to combine the strengths of Unity and CUDA engines. Can I do it in any way? For example maybe CUDA provides the ability to write some code, and then placing it in dll, so that other applications can use it automatically?

Yes, you can create a DLL that contains GPU-accelerated functions. You can find a lot more information about it by googling “cuda dll”