Right, I see. That's not really possible imo. For things like mlops, sure. But model development, selection, evaluation? From what I've seen, it's exactly when engineers reach for standard tools without giving thought to how it applies to the given problem that they do a bad job.