This onsite interview includes five rounds (late Oct) and I failed at the hire committee. The position is a machine learning software engineer. After first round behavior questions, which were quite general and compared to Amazon also not so agressive, two machine learning interviews and two code sessions. Due to NDA, I only describe the main ideas behind those coding interviews. In fact, some problems are quite similar to the leetcode problems. One question is related to maintaining a top k heap and continously insert and delete operation. Also I got a question similar to square root, but with different requirements, and the basical idea is binary and recursive way to approxmiate with additional accuracy range. I also got problems quite close to word pattern matching, and I think medium plus level.
For machine learning, design a system like linkedin to find the top rated candidates, and also lots of details about the network learning and modeling, but quite general.