Well, after doing some research on the existing software I feel it's going to be much easier to create my own software, I still need to do some more analyzing but writing solvers for the FE is really easy seeing as there are really on 3 or 4 useful elements: beam, 3 and 6-node triangle (I know 3-node triangles suck but they are just so easy to implement), 4 and 8-node square and 8 and 20-node bricks. The hardest part that I can think of is creating a mesh generator, which I have no idea how to, but I'm going to go on a limb here and say it's just implementing a well established procedure.
The issue with importing outside data is really not an issue seeing as most are standards like IGES and STEP files, which are actually plain ASCII files. There's also a library out there for MED files which can be used.
I also want to remind that what I'm looking at implementing right now is just plain FE, not a complete 3D modeling framework as this is without a doubt a huge undertaking, hopefully someone else will find my work with FE interesting and write a CAD/3D modeler library/framework. A pure FE framework wouldn't even need a post-processor as it wouldn't have to know how to represent it, just how to calculate the data from already existing node/elements.