There are four steps. First, you need to complete a work sample where you need to fit a linear model and write up your entire solution as a report. Next, they will schedule a phone screen with you to talk about the work sample, and ask you a coding question as well as a machine learning question. Very easy questions, no preparations needed. Then they will invite you onsite to meet with the research team. There were 5 rounds. For each round they will ask you coding problems, probability problems, machine learning/statistics problems as well as your projects. As a final step, they will ask you to provide as many references as possible, including your advisor and your manager (if you are working in industry). As other interviewees have described earlier, the entire process can be best described as long and frustrating. They provide absolutely no feedback as to why they reject me. They simply say that there isn't a match. After all these you finally realized that my background is not a good fit? This is simply ridiculous and unprofessional. Why would you make reference calls if the candidate's background does not match the job? They are looking for machine learning experts with strong mathematical ability and professional software development skills. Honestly, anyone that has all three skills will not even consider working for them. Most of the research team consists of postdocs and previous faculty members. If you do not have a ton of machine learning related publications, or if you didn't graduate from Berkeley, don't even bother.

My interview process started with a phone screening. The interviewer was another PhD in a similar area, and he asked some statistics and math questions. During the conversation, I got the impression that he did not particularly like the job himself, so I asked him about it directly. He said that the role was not very flexible, and that new hires were initially assigned to groups somewhat randomly. In his view, some people ended up in groups that were clearly not a good fit for them. The next day, I was invited to the first part of the virtual onsite, which consisted of two rounds: one in statistics and one in machine learning. The statistics round was conducted by a statistics PhD. The ML round was conducted by someone hired for a background similar to mine, so although he did not seem to know much more about the broader process, the conversation happened to align quite well with my own area.