You need to consider, that the transformations are matricies (4 vectors) , but matrix-scalar multiplication is commutative and matrix multiplication distributive, therefore you can rewrite the code for better performance.
Ashaman73, for optimization purposes could I do the vertex multiplication only once like this:?
Better to take a known export format like FBX/collada and write a converter between export file and your binary file. This way you can easily support other modelling tools like blender or maya.
I'm thinking about writing a max script to export that information into a proprietary format and my engine would load that. That way I would not need to deal with other formats.
The bone interpolation should be done on the CPU, the vertex transformation on the GPU.
At this point I've got the animation of the skeleton loaded into my engine. I just need to animate my object based on that. With this approach I would not be able to benefit from VBO if I decide to perform caculations on the CPU, right?