This is exactly it, NVIDIA is continuously “understaffed” and many ICs end up splitting responsibilities across different projects at different places in the stack. In stark contrast to many other large companies where an SWE team might have several entire other teams of 10+ engineers just to support the tools, build systems, test runners, etc.

Engineering groups are more fluid and individual contributors get moved around a lot (relative to other companies) according to demand. And management groups don’t really have “fiefdoms” the same way they can at huge companies… possibly due to the controversial matrix management employed here.