That was my first thought too. I'm about to finish my masters program and I spent the first half thinking that it was just hyperparameter tuning, until I sat down and learned the math and theory. Now it's so much more interesting and explainable. That random tuning is now much more calibrated from experience and understanding the theory. (As of now), I could easily make a career out of this, because it's not random and simple optimization. Plus, the field is so hot right now, that it's unreasonable to assume that what data scientists do now is what they will do in 5, 10, or 20 years