I interviewed for a Data Science 2 role. First went through a basic screening with the recruiter, then got a link in a few weeks to record timed video responses to 6 questions over 2hrs ranging from technical to theoretical problems.
Got invited to the onsite interview and had to cut it short because it was so bad. I went in very optimistic, but ended up having to answer a series of irrelevant questions like 'how many layers does BERT network have', 'list all classification models', 'explain xgboost' (I've never worked with xgboost) etc.
People who ask questions like these in interviews shouldn't be doing this job. It shows lack of creativity and insecurity about their own intelligence. After 35 minutes of these questions in the second of 4 interviews I had seen enough.