One of our simplest “screening” questions for DS roles at my company is: “your model is 100% accurate. How do you feel?” If the answer is anything other than deep skepticism (Data leakage, trivial dataset etc), it’s a big red flag