I think you're right, they're pretty wishy-washy, but I'd define data engineer as someone who builds systems that make quilty, useable data available (i.e. anything from building ETL pipelines to productionizing models), vs a "data scientist" which I'd probably describe as doing more one-off in-house research type work.

I suspect a lot of "data scientists" end up being "people who write tableau reports for other people" and/or "people who manage an ugly pile of python data processing scripts to make the data-spice flow"

In my experience the plumbing is a lot more work [requires more man-hours] than the interesting visualizations, and I think some organizations do a good job of supporting a few scientists with a robust engineering staff, while others hire the scientists because they want the fruit, but forgot to plant the fruit tree.