Seriously, don't underestimate the classics. A lot of people assume they have intractably large problems when they honestly don't. Just because a few TB was "Big Data" that required a fancy Map Reduce cluster 20+ years ago, that doesn't mean you still need to think about every problem in hyperscale terms. You can grep through a few terabytes of files much faster than you can research and propose a fancier solution, let alone make use of it.

And I say that as somebody who has spent a lot of my career working on and around big clustered and distributed filesystems. There certainly are cases where you need complex stuff. But always start with the simple things.