It was pretty similar to what others have said here except the coding test (I will discuss that in a minute). So, the first round was the usual set of screening questions by an HR rep: (1) tell me about yourself ("your story"); (2) what kind of ML experience do you have; and so forth. At this stage, make sure you throw in as many ML buzzwords as might be applicable to you. At the end, the HR rep also asked me if I had questions for them. The second round was a technical interview with a data scientist there. They asked me go in-depth into one of the ML projects that I had worked on. So basically I had to give a step-by-step description of the project with the interviewer asking me questions in between. To prepare for this stage, I would make sure to brush up on past projects listed on the resume (although they did let me choose which one I wanted to talk about). They also asked me questions about my related experience and background. Following this was a coding challenge where you will have 8 hours to complete a classification task. Some of the earlier posts suggest that this is a dataset with 8000 observations or so. This might have been true when they took the test, but the dataset is now much larger. It was closer to 200,000 examples and almost 500 features. So obviously make sure to use a computer with decent memory and brush up on your dimensionality reduction techniques (PCA etc) before starting the test (as it is timed -- 8 hours max). The dataset is a "real world" one in that it has missing data, erratic categories, and so forth so be prepared for some data cleaning. Other than that, the classification task itself is pretty standard and the problem is more or less what others have mentioned. Anyway, this is where the story unfortunately ends as I was not selected for the next round of interviews. All in all, the interview process was pretty standard and laid back and the individuals I interacted with seemed nice.