I had similar experience in some big companies.

Bombed the leetcode, but found an opportunity to show-case my (fairly cool) project code during technical interview. Asking the guy questions, he confused feature importance with feature selection, couldn't answer about a baseline model (They had a black-box without one), and a bunch of other things. When I said "I kind of prepared for pandas + SQL more", said "We expect you to know those things". I guess they expect me to know how to use pandas and SQL but not python for crappy leetcode questions.

The truth is, most companies/ml departments have no idea what they want or should be doing. Good luck to that head of ML team, because I was glad I wasn't selected, with such great interview and ML skills it's a bullet dodged.