5 Things I Did to Prepare for An Applied Scientist Interview at Amazon
“How did you prepare for your interview with Amazon?”, “How did you prepare for an applied scientist or machine learning scientist interview?”
These have been recurring questions that I receive among my friends and to be frank, there aren’t many resources out there on a preparation plan for applied scientist (at least during the time I was preparing for the interview), unlike software developers job roles. Therefore, I am writing this post to share my experience and to help any folks out there who will be interviewing similar roles. The experiences that I’m about to share will be based on my experience interviewing for an L5 applied scientist role at Amazon. However, I had also employed most of the the following strategies when I was interviewing with other companies, and I believe it will be applicable for similar roles at different companies.
1. The basics of machine learning
We all know that the longer we spend our time in our current job dealing with complex problems, we tend to forget the basics. One of the things that I did was to refresh my memory about the basics of machine learning and the theory behind it. Things such as bias and variance tradeoff, regularization techniques, gradient boosting algorithm, linear regression are fundamentals of machine learning. I took the Statistical Learning course from edX (taught by T. Hastie and R. Tibshirani) to keep my mind fresh with high level knowledge of the fundamentals. During the course, I realized topics which I was rusty and I did extra study on these topics. For example, I have always known about the bias-variance tradeoff, but the details of the derivation of it was never clear to me. Similarly, why L1 regularization favors sparse parameters? See these posts if you would like to learn more (bias-variance tradeoff, sparsity of L1).
Remember to keep notes so that you can refer to it whenever you need a quick refresh before any interview!
2. Understand all topics listed in your resume
Good resume lands you the opportunity for an interview, make sure you are able to explain ALL topics that you listed in your resume during an interview! If you used an object detection network in one of your projects, you should spend time understanding the structure of the neural network, the techniques employed, metrics, results etc. Even better, you can study more about different networks which achieve the same task, understand the pros and cons of each network. Then, you are equipped with knowledge on why you have selected a particular network as the algorithm for your project.
Remember to also take notes for a quick refresh!
3. Leetcode and mock interview
A requirement for an applied scientist role in Amazon is an L4 SDE coding requirement. I had to admit, this section of the interview was the section that I was most terrified of. I am pretty comfortable with general coding, ML pipelines and infrastructure, but I didn’t feel comfortable with leetcode-style interview. Anyways, to prepare for it, I had to feed myself with data structure and algorithm knowledge. Fortunately, during my undergraduate, I did a minor in computer science and I took several related courses, not too much of a hassle!
Theory itself wasn’t enough for the preparation, I had to equip myself with practical applications. I got a Leetcode subscription and started to practice the questions almost every day. Initially, I picked a couple of easy questions to start with, getting myself tuned into the “coding” mode. Then, I tried to tackle about 2–3 medium level questions every other day.
I will try to give a few tips here if you have a leet-code style coding section in your job interview.
- First of all, try to understand the question fully. Ask any clarifying questions if it’s unclear and take your time to do this.
- There is always going to be an example given to you. Try to walk through the example with the interviewer and explain your thought process using the example. You will almost always get a solution (maybe not optimal) when you walk through the example.
- Explain the time and space complexity of your first solution
- Sometimes, you may want to code your initial solution to make sure you have a working solution in the case that you’re unable to figure out a better solution. Ask the interviewer about it.
- Optimize your solution and take hints from the interviewer when given. Often, the interviewers will guide you through the process and they are always helpful.
- Finally, code your final solution and explain the complexity of your solution.
Before my actual interview, I had 1/2 mock interviews with my friend. It really did help a lot! Also, you should try your best to speak your mind throughout the interview process so that the interviewer understands your thought process and he/she would be able to provide the appropriate help when needed.
4. Behavioural questions
There will always be some sort of behavioural questions in your interviews. In Amazon, it is the leadership principles. Please make sure that you are well-prepared for these questions! To tackle such questions, you can utilize the STAR method to answer the questions. For example, let’s say the question was “What improvements have you made at your current company?”, a sample question from here.
- Situation(S): What was the situation and the problem you are trying to solve? E.g. Company X has a strict rule that only dogs are allowed to guard the door. We hired a Shiba Inu (Rocky) to screen all animals that came for the job. If he barks twice, it means the animal is a dog. Otherwise, it is not a dog. As Rocky has seen himself in the mirror for too long, he only approves any animal that looks like him (White Shiba). While Rocky is cute and hardworking, the company does not think that he’s a good fit for the job and wanted to create a new solution.
- Task (T): What were your tasks in specific? E.g. I was given a task to identify dogs.
- Actions (A): What actions did you take to complete the task? E.g. With a machine learning background, I collected images of different dog breeds and trained a ResNet model to classify dog images. Whenever there’s an animal who is interested in the position, I took a picture of it and input it into my model. My model returns a probability from 0 to 1 whether the animal is a dog.
- Results (R): What were the results? Are there any quantitative results? E.g. In the hold out test set, I compared my model to Rocky’s classification rule and my model has an accuracy of 98% and Rocky’s accuracy was only 50%. Moreover, due to the increase in accuracy, the incidents of cats lurking around and stealing snacks from our cookie jar have decreased by 20%, which further saves the cost of the company by 10%. While Rocky might be sad as he lost his job, he is definitely happy with an extra cookie in the jar and has been given a new role as a cookie jar guardian.
For my Amazon interview, I was prepared for all leadership principles and prepared 3 examples for each of the principle. Also, keep practicing the STAR method for each of the examples you prepared, whether you are speaking to yourself in the mirror or having several mock interviews. Either way helps!
5. Keep calm and meditate
Interviewing for a new job is never easy. It is intense and overwhelming. You might be worried that you are not well prepared, or you are anxious about the interview itself, or you are anxious about the 100 little things that might appear during the interview. One thing that I find useful is to meditate! Whenever I feel overwhelmed, I will take a 15 min break a day to meditate. When I feel anxious before an interview, I will take a few minutes off just to meditate and calm my “monkey” mind. Try it out and you will notice the difference! At the end, you will find that these interviews are not that scary as it seems, and it could sometimes be fun especially when you share your experiences with an experienced scientist and both of you have a common topic.
Last but not least, revise the notes that you have jotted down throughout your preparation process. You do not want to waste all the time you have spent studying! I hope these tips are helpful, and let me know in the comments about any other useful tips.
