The process began with a recruiter reaching out to me to do a take-home exercise with three parts: some medium-difficulty SQL queries, a qualitative question on metric evaluation and experimental design, and an applied predictive modeling exercise for which you needed to submit code. After that I did a 30 minute Uber-related case study over the phone with a data science manager before the final on-site. The on-site was comprised of four interviews: one with a PM which was kind of a combination of background and explaining technical stuff to non-technical people, two data science related case studies about Uber problems that involved both experimental design and predictive modeling (with some whiteboard math/brainstorming), and one solely focused on metrics and experimental design (which they emphasize a lot if you couldn't already tell). The take-home coding assignment was probably the best / least annoying of the 4 I did during recruiting, and all the interviewers were really, really sharp. Only negative was that the onsite was quite rushed and just a 4 hour block straight through with not much time to collect one's thoughts after each interview.