Amazon

I applied for the Applied Scientist Intern role at Amazon through the off-campus process. The first stage was an online assessment on HackerRank. It had two relatively straightforward DSA problems along with a short set of Leadership Principles questions. The coding part was easier than what I had expected from an Amazon SDE assessment, so I concentrated on getting both solutions correct rather than trying to overcomplicate them. After clearing the assessment, I received the interview invitation. The first technical interview was almost entirely about problem solving. I was given two DSA questions, one involving the next greater element pattern and another involving traversal of a binary tree. I explained the approach before starting the implementation, and the interviewer kept asking about complexity and edge cases. I completely coded the first problem, while on the second one I was able to explain the approach and partially implement it. The interviewer was more interested in how I reasoned about the problem than in simply seeing finished code. The second interview was very different. It was specifically focused on Machine Learning, and this was where the role became much more specialized. We started with my ML-related projects. I had to explain why I chose particular models, how I prepared the data, how I evaluated the results, and what I would change if the model did not perform well. The interviewer then moved into ML fundamentals. I was asked about linear regression, logistic regression, KNN, decision trees, and random forests. The questions were not limited to definitions. For example, I had to explain why a particular algorithm would be appropriate in a given situation and what could cause a model to overfit. Decision trees took up a surprisingly large portion of the discussion, with several follow-up questions based on my previous answers. I was comfortable with the basic concepts, but I struggled when the interviewer started going deeper into the theoretical reasoning behind some of them. I realized that knowing an algorithm at the implementation level was not enough for this role. The process ended after the ML round, and I was not selected. My biggest takeaway was that an Applied Scientist interview sits somewhere between a software-engineering interview and a dedicated ML interview. Preparing only DSA would not have been enough; I needed much stronger depth in the mathematical and theoretical side of ML as well.