Sorry, but string prefix search over a few hundred million entries is something you can do with the same performance using just postgres on a single server with just a few hours of dev time.
I've done it before, it's not as impressive as it seems.
With trigrams you can even do precise substring search on this scale with good performance.
I'm sorry, but that's wild. That's like, not even an easy engineering problem to solve necessarily, given their size!