A shorter version of how I described it above is that Data Engineers intentionally decouple logic from data so that ETL processes can be managed at scale (not necessarily talking about Big Data when I say scale. Sometimes lots of small ETL processes are just as difficult to manage as Big Data).

Don't think he is that wrong. Putting aside Spark, the remaining terms are pretty broad.

Data Lakes: Is this our new fancy term describing "data". So what distinguishes "data lakes" from "data"?

ETL: I would guess 95% of application programs take input, parse it (extract), do some data wrangling (transform) and save the result somewhere else (load).

Validation: Again a broad term. Do you mean validation of statistical models? Without validation your predictions are worthless so I guess it is a standard thing to do if you want to do any kind of machine learning.

Schema Management and data catalogs: Standard DB stuff I would say.

We just like to define new job descriptions. It's the same with DevOps, which seems to be the new term for System Administrator.