Excellent, thanks everyone for the advice! I think I could easily produce a barebones beta and just push it to grad students (the intended user) for free, see what happens. If people adopt it, perhaps I can investigate adding bells/whistles as you mentioned.

How big of an audience/revenue stream should a startup shoot for? I really have no idea. Just for fun, I ran some quick numbers:

A 10k yearly fee works out to 0.04% of MIT's library budget (26m), and 0.27% of my alma mater's budget (3.6m). If 10% of 4-year universities adopted the service (~230 colleges) that totals 2.3m yearly.

Obviously that is a really simplified example. I imagine larger schools would pay more than smaller schools, 10k is probably a silly number, ten percent adoption is probably overly-optimistic, etc. But I suppose it puts everything into rough context.