First round:

Rescheduled twice, pretty last minute. Friendly interviewer. Two easy to medium LeetCode questions; no ML questions. Overall experience was alright. However, the interview was conducted in Chinese, which was quite unexpected as it was never specified in the job description. Suspected favoritism towards Chinese-speaking students.

Second round (final round):

This is where it got pretty messed up. Rescheduled once last minute due to interviewer unavailability. The interviewer started the interview without first introducing himself or mentioning his name, and did not ask me to introduce myself either. He proceeded to two easy to medium LeetCode questions afterwards, which I answered perfectly. During the coding questions, he provided little to no insight and seemed impatient. He did not even understand the formula for dot product in matrix notation, which was a surprise given he graduated from Oxford CS.

He proceeded to ML-related questions in the second half. He grilled me on recitation-style questions (like the formula for a specific loss function and vague questions like 'what is the most popular regularizer in NN?'). I had to admit that I underprepared for the basic ML questions, though those are memorization questions that could be looked up in three seconds. This probably resulted in rejection.

LC + ML questions, no behavioral.