[maximente]

deliver value to a market willing to pay for it.

most technologists including ML engineers are more interested in how the latest, greatest kernel provides a 0.04 higher F score on some academic facial recognition task than they are understanding problem spaces, understanding what problem spaces would benefit from "ML" (read: data + statistics), then providing a solution that people value enough in order to pay for (this includes sales, marketing, etc.).

or just work for a FAANG, probably easier and over time a wiser investment