4 rounds. Two leetcode coding. One Pytorch coding. Ask about ML questions and your own research. Need to be familiar with relevant papers in your field, and also the fundamental implements in Pytorch.

There was 6 interviews over 7 weeks. No clear schedule was provided by the recruiter and everything seemed organized on-the-fly with interviews every 2 weeks. Content of interviews is never clearly announced (e.g., only "ML interview") and thus you cannot prepare accordingly nor request any accommodations (e.g., please, no live coding). No ML system design, no AI design, no research talk, just ML culture questions and ML coding (pytorch). Also note that "casual chat" is their phrase for behavioral interview. Everyone was very friendly and easy to talk to. As a research scientist, I performed very well. But after, 3 live coding interviews (yes, 3!) and 2 ML culture interviews, it appeared that they were searching for a ML engineer... At some point I was bored by their coding questions, especially when they ask me the same questions twice during different interview. All questions were easy, it is only about your motivation to answer this kind of questions and whether you are OK with live coding. No one asked me anything about research: methodology, ethics, experiment management, evaluation, etc. most are on the job ad but never asked in interviews. The job ad only mentioned "basic coding skills" so I was not expecting this kind of interview. This was very unusual, especially the pytorch coding (the job ad did not mention anything about ML framework) and it really seems that the company (at least the people who organize the interviews) don't really know what is a research scientist. Pure waste of our time. This should never happen. I do hope their team will get a true research scientist and not yet another ML engineer. They really need one.