Classic stuff about your experience and some examples of relevant projects (for the phone interview) maybe a little bit of model stuff like why you chose specific models. They didn't seem happy with my answer of "I choose the easiest thing to implement first" and carried on trying to wrangle a more data scientist-y answer like "A decision tree seemed to mathematically fit the feature space better", which is fine if you want to make sure people understand model basics then ask it directly rather than try to direct them into a weird, non-truthful narrative of what happened on the job. In person, 3 hour interview all on model specifics. Asked a lot about maths. I think these are all fine, depending on the role you're applying for. Personally, I'm more of a full stack DS person and by focusing so much on the models and the maths behind the models, it was an odd interview for a role that I assume ultimately was applied DS rather than research DS. Or maybe it truly was a research based role and the questions were appropriate. The fact that I still can't tell says a lot about how unclear the role description was and about the confusion from the hiring manager about what they wanted. That's a red flag.