I applied online (Machine Learning Scientist at AWS) and contacted by a recruiter after a month or so. Technical phone screening is scheduled subsequently for one week after. In the meantime, I studied for Machine Learning, Probability & Statistics, Programming questions. The interviewer was a nice person, but was a bit confused himself; and sometimes was not so clear in his questions. The questions were all Machine Learning questions, no Statistics/Probability, or Programming (even though there was a shared screen for coding). He asked about a project I've done on Machine Learning and questions mostly evolved from my answers (careful about what you say!): the difficulty level mostly depends on the knowledge you already have. The questions were overall easy and conceptual, and required a general knowledge on various topics.

Call with a recruiter followed by technical interview. The technical interview covered broad Machine learning questions. Then the coding part. The problem required prior knowledge of a calculus theorem, which in my opinion was not fair.