In October 2023, Amazon India visited our college for full-time recruitment. I will be sharing my interview experience for the Applied Scientist role. The process is very rigorous, initial shortlisting is based on a CV followed by 5 interview rounds.
- Data Structures and Algorithms + Resume
- Machine Learning Breadth
- Machine Learning Depth
- Scalable Machine Learning
- Machine Learning Applications/Managerial Round
1. Data Structures and Algorithms + Resume
The interview kicked off with a friendly introduction from the interviewer, allowing me to get a sense of their background. Following that, it was my turn to provide an introduction about myself, which spanned about 4-5 minutes. Once that initial exchange was complete, we dived straight into the coding segment of the interview. They provided me with a link to an online platform where I would be engaging in live coding. While I don’t recall the exact coding question, I do remember that it resembled a smaller variation of a common coding challenge.
- 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.
- Reverse a linkedlist in size of K
Above both questions, I am able to solve them within the time period, in the first question I was stuck but the interviewer hints me to think in the right direction, So advise is to try to give my best while solving question and explain the interviewer in-between what you are thinking will help you.
The coding question took up around 40-45 minutes of our time. Once I completed that, the interviewer delved into my CV, specifically focusing on the internship I had done during the summer. They asked for a detailed breakdown of my responsibilities and the kinds of problems I tackled. I provided a comprehensive explanation of my experiences. Additionally, they inquired about the metrics I utilized, the loss function I employed, and the reasoning behind my choice of a particular model over others. I provided thorough explanations for all these aspects. Throughout our conversation, they were interested in understanding the impact of my work on the company’s metrics, and I was able to convey this insight effectively. This seemed to impress them, and with that, the first round of the interview concluded.
2. Machine Learning Breadth
This round encompassed a comprehensive exploration of machine learning, covering the underlying mathematics of algorithms. Additionally, it delved into the realm of deep learning, encompassing breadth in terms of understanding and applying mathematical concepts. The questions asked during the interview are as follows:
- Explain the decision tree.
- Explain Random forest.
- How is Random Forest different from Decision Tree ? and why is RF better? (Random forest uses multiple DT, high variance is high, each tree depth is high, we are combining multiple DT take advantage that each learns differently expect of data, trees are uncorrelated and we give a portion of data and feature to each model, he is not satisfied and repeating the same question even didn’t give the hint)
- What is regularization, is dropout a regularization, and why?
- Write the equation above and prove I tried. But he his not satisfied and then moved on to the next question
- Explain Dropout and how weights are used at inference time.
- Do we need scaling while inputting in NN? if yes in which case?
- You have 2 model’s fit() and its eval() functions, train and test data how you will check which model performs better and which model is overfitting? train error low and test error high does not answer, lets say you have completely new data in the test.
- What is regularization, is dropout a regularization, and why?
- Asked me about the K-means algorithm.
After that, I didn’t get a call for the next round.
Summary:
Despite my best efforts and thorough explanations, I received a rejection for the position. While disappointing, the experience provided me with valuable insights into Amazon’s hiring process and the expectations they have for the role of an Applied Scientist. So, in an interview, you must have breadth as well as depth knowledge and also need to have skills that come into Amazon leadership principles.
Verdict: Rejected