I took a total of 8 interviews for this position: 2 screens and 6 on-site interviews. One screening was for software engineering with general questions regarding classes, functions, data types and a coding problem. The programming language is at your choice, but you need to write actual, runnable code. The second screening was with data science, again with general knowledge regarding how decision trees work, backpropagation in neural networks and dimensionality reduction using SVD. One on-site interview was with the manager and it was all about the Leadership principles. You can find examples of question online, but if I were to give just one advice for the Amazon interview is this: take seriously the Leadership Principles (LPs)! Probably Amazon is one of the fewest companies in the world where the HR interview is more important than the tech interview. Prepare 2 situations for every LP and be prepared to be challenged on your answers and asked to dive deep on the situation and most importantly on YOUR contribution. Another interview was a presentation that I had to hold with a subject on my choice. I presented my bachelor thesis, it's a very common topic. I received a few questions, but nothing too complicated. There was another software engineering interview where we again touched on the fundaments and I received the 2sum problem. I gave the dictionary solution. He got deep with precision of the numbers in the memory. We then talked heap sort vs merge sort and binary search. Next, I had 3 data science interviews. First, "breadth" with general discussions about bias-variance, activation functions in NNs, t-test, bonferoni correction, transfer learning, l2 vs l1 regularization, bagging vs boosting. The second was "applied" where I got a an open problem: find anomalies in sequences of events. I approached it with RNNs and tried to predict the next event in the sequence and flagged as anomaly when the true event was not in top 5 in softmax. We then dive deep on how GRU cells look like, vanishing and exploding gradients and different loss functions (cross entropy vs KL). The last interview I don't remember, sorry.