By Murtuza Bohra | Updated December 11, 2020

This blog is useful for you, if you are preparing for applied scientist position at Amazon. In this blog, I have listed all the questions asked in different interview rounds. Applied scientist in Amazon are expected to have good machine learning knowledge as well as programming skills. The requirement include past work expecience of dealing with large data and machine learning, and if you have any publication that will be a plus point. You can continue reading, if you are curious about what kinds of questions are asked in the applied scientist interview. This blog is purely based on my personal interview experience.

Interview Round-1

• Find the largest sub-string in lexicographic order for a given string. For example, if the given string is "ababaa", then the largest sub-string in lexicographic order will be "babaa". You can find the solution here.


• Find the largest sub-sequence in lexicographic order for a given string. For example, if the given string is "mississippi" then the output will be "ssssppi". You can find the solutin here.


• How does neural network (NN) works, explain as if I (the interviewer) have no idea about it. Then follow up question was to explain backpropogation.


• While explaining NN I had mentioned about dropout, so the next question was "what is dropout and why is it used".


• Can you explain how SVM works, then the follow up was to explain the mathematical formulation of SVM.


• Above set of questions were to judge the theoretical undestanding of ML concepts. Finally, inteviewer asked me about one of the project mentioned in my resume.

Interview Round-2

• This round was more focused towards my resume to better understand my work experience. Started with asking me about the experience of the longest project I have worked on. To which I chose to discuss about my recent research project. All the questions were straight, but they were wisely arranged in an order. Following were the questions.


• what were the three biggest challenges you faced in this project.


• Followed by couple of questions around the project and related concepts, may be, to better understand my project.


• Then, for each of the challenge I mentioned in the previous question, he asked me to think of three different approaches in order to addreess these challenges.


• Then, he asked me to explain him, how the approach I had used originally in the project is better than these alternate approaches.

Interview Round-3

• Again started with some questions around my resume to understand my area of interests.


• Explain what are generative model in machine learning.


• What is the difference between generative and discriminative models.


• I had given examples of Naive Bayes and Bayesian Network for generative machine learning methods. So he askes me to explain these methods.


• Then, the interviewer asked some Adhoc problems in the context of amazon e-commerce. i.e You are given a large list of products, and each product is described with textual description and a set of product images. Now if you are given an arbitrary product image, how can you identify which product this image belongs to.


• At the moment, top of my head, I could think of two different approaches for the problem which draw rest of the interview. I told him one approach using classical computer vision based feature extraction and learining a SVM based product image classfication. And the second one was based on deep neural network (Siamese Network) to distinguish between two different product images.


• Then he asked about the loss function and how to choose one which is more suitable for the above problem.

Interview Round-4

• By now I was sure that the interview will start with my resume. As expected, interviewer asked me to explain a project which I have not talked in previous interview rounds.


• I thought interviewer might have access to the feedback of my previous interview rounds, so I honestly chose project which I hadn't discussed before with other interviewers. Then, he asked me about an incidence when you were stuck in the project and how did you find a way to solve it.


• Having discussed about the project, he asked to tell my history in 2 mins. Though it was a strange question, you can actually use this opportunity to convence the interviewer by aligning your past interest and experiences with the job profile (applied scientist in this case).


• Interveiw so far was kind of non-technical, now finally he asked an me to choose a programming language and gave me a coding question.


• Given two dates (DDMMYY), find whether they are 1-month apart. First of all, he was expecting me to immediately ask a question which I could sense from his expression. Generally, in such tensed situation we forget to read the interviewer, but It was a lucky day for me. I asked him "what defines 1-month", which is exactly the question he was expecting. This is because every month has different number of days (also need to account for leap year) and if the given date starts in the middle of the month then how many days do I count for one month.


• Finally the interview ended and he asked if I have any questions for him.