Several months back I was figuring out how to use pip and virtualenv, and generally having a pretty good time with my python development experience. However, I kept running into interesting code examples and projects that required matplotlib and the numpy/scipy stack. Since I've also been trying to get familiar with mathematical modeling, analysis and visualization techniques, these are indispensable dependencies.
I could (and had) play with them "raw", but ran into problems when I tried to enjoy the control and isolation afforded by virtualenv. Last week, though, in one of my periodic attempts to comb through dependency installations, I stumbled on this gist, and it cleared up all my problems. I haven't delved into whatever reasons may have existed for my previous difficulties, but if you're having trouble, maybe it will help you. All I did was to set up a new virtual env and pip-install my way through the list. It... just worked, for the most part (thanks, fyears!), although I did have to update some fortran-compiler stuff that numpy needed.