How to find a team with specialized technical depth?

First, I recommend going through this other discussion from a Meta engineer about team selection, where I talk about all the avenues you should use to evaluate a team.

To start with the obvious, I think you can just ask. Literally just go to the tech lead and senior/staff engineers on the team and ask, "Will this team expose me to a lot of technically deep ML problems like neural architecture?

But of course, folks, especially managers, may not be 100% honest when giving you the scoop on their team. To get around that, here are the information sources I recommend digging into to get the more "real" picture of the team:

1. Team insights - In particular, go through the diffs of the E5s and E6s on the team. If they're exhibiting technical specialist behavior, you will be able to see it in their code. Go through individual diffs and you should be able to grok from their context how technically challenging they are. Good engineers will often times have something in the diff summary like "Solving this was very hard, but here's the approach I ended up on..." - That's a good indicator of technical depth. And of course, volume should be decent - There are teams where the senior/staff engineers will do <25 diffs a half as the work is mainly XFN oriented as opposed to technical depth oriented.
2. System design docs - This will come in a combination of Workplace posts, Quips, Google Docs, and tasks. Similar to the diffs, there should be a good amount of these, and they should be thorough and long if the team truly has technical depth.