EliasDB – A graph-based database

I think the distinction isn't so much between "student/hobby projects" and everything else, so much as it is between "projects that go through the motions overly-formally to learn the motions, rather than to achieve the goal" (i.e. the result of doing a code kata) and projects that are created with intent to put them into—at least hobbyist—production. Even quickly dashed off scripts are useful to others; while perfected, deliberate homework assignments usually aren't.

Partitioning the search results of sites like GitHub based on such a factor might be useful.

You're right, though, that people wouldn't know well-enough at the time to self-report this "is practically-applicable rather than an academic exercise" property. It'd be more useful as a fact derived from social-bookmarking/tagging of repos, I think.

GitHub already has a pretty good signal for practicality in its "stars"; people don't star things that aren't useful to them in some sense. You could maybe add some other explicit signal (maybe a "hide" button in the search results) to serve as a negative signal.

Better yet, though, you could build some sort of "BlackBoard for CS" education-workflow site on top of GitHub, and use its positive signals (this is academic) as the negative signal for GitHub itself.

(As an aside: I've always wondered why more sites don't use this approach. It's way easier to incentivize people to tag something they want vs. something they don't. Build a second frontend for your site that serves the user-base you don't want on your main site, and serves them better—and then have that frontend add the ghetto-tag to everything it passes to the backend.)