It’s not a given in a big company that they’ve got their data-driven infra/culture figured out, there’s always people coming in and out and some people need to learn it. =)

But assuming there’s good practices in place, a junior scientist can learn these things by paying close attention to the processes in place. Culture often manifests in semi-formalized processes, like if they’re in the habit of setting metrics before launch, and missing instrumentation can literally hard block a launch. They won’t KNOW that they’ve picked these skills up, but they’ll feel very frustrated if they move to a team or company that doesn’t have it. Enough that they could champion it.

The alternative is to find a startup-y team within BigCorp, like a group that does early new feature work where things are very ambiguous. Either they’ll have good practices already, or they won’t and you can help build them.