For additional context, we are currently using problems which have test cases, and it feels as if it doesn't correctly reflect the day-to-day of a data scientist. In fact, many interviewers pass a candidate iff they pass the test cases.

I'd love more open-ended problems.