I mean, someone in the 'maker community' has a high probability of having some difficulty, but my 12 year old cousin could build the driver on protoboard, no issues, and it's not like he's worked through 2nd order diff non-homogeneous eq's ;). That's the only 'strange' thing about it is the voltages as you mentioned (well that and the somewhat rare flat-flex ribbon used). It needs a jellybean boost converter to get up to the +/- 25V @ a few tens of mA. This guy[1] did it for what looks to be under 10 bucks in parts. Some LT standard (not as standard as the LM317 or 7805, but close) will switch both rails no sweat, all day, then use 2 standard LDOs will get get your -25 up to -15 and regulate your +25 down to +15. (I don't like his value choices for those caps around U4 and U5, but the rest of the design is so well done and I'd imagine he used SPICE and tested every node, so I'm going to just defer to his design, since I'm not an EE haha). Everything else are jellybeans you'll have lying around (except maybe those inductors).
The only thing you have to watch out for is quiescent current, not only for the obvious reasons (leakage == bad), but also because even a minimal amount of voltage drop 'across the FET' will fade out the image, so its double bad. (Each pixel with the characteristics of an N channel MOSFET.) Acting as the source on the FET -- continuously driving "on" for each "frame render" (it's just latched in from a shift register), and the horizontal pixels corresponds to the vertical pixels (which is just a shift register pushing data for each render).
If you have issues with that guy's build, http://spritesmods.com/?art=einkdisplay this dude consistently has some of the best documented hardware hacks ever. Those two sources in aggregate and a Saturday should be all you need for a one off.
[1] http://essentialscrap.com/eink/schematics.png