Amazon SDE-2 | Chennai | Rejected
Summary
I interviewed for an SDE-2 role at Amazon in Chennai, which ultimately resulted in a rejection. I performed well in the DSA and LLD rounds, but faced difficulties with a Machine Learning-heavy HLD question and a challenging variant of the Longest Common Subsequence problem during the Bar Raiser round, leading to the rejection.
Full Experience
The recruiter initially reached out to me via call and followed up with an email regarding the interviews, sharing the Online Assessment links. I completed the OA in the last week of June, and after selection, the scheduling call happened in early July. I was informed to expect 4 virtual interview rounds.
Round 1: DSA
My first round was focused on Data Structures & Algorithms with an interviewer from the US (8 YOE). After a quick 5-minute introduction, I was given the LRU Cache problem. This was quite unexpected, but luckily, LRU is one of my favorite problems, and I've solved it over 20 times. I explained my approach and solved it within 10-15 minutes. The remainder of the round involved two Leadership Principle questions and a detailed discussion about my current role and projects. The feedback for this round was 'Inclined'.
Round 2: LLD
The second round, focusing on Low-Level Design, was with an interviewer from India (13 YOE). After a brief 4-minute introduction, I was tasked with designing a Bus Ticket Booking System. I was given 6 requirements to fulfill for the design, covering aspects like service design and concurrency booking. I had prepared primarily for LLD focusing on UML diagrams, class skeletons, or coding implementations. However, the interviewer asked me to design the API endpoints and database schema, which turned it into a mix of database and microservices design. I had to rely entirely on my experience and reflexes rather than direct preparation. It felt more like an HLD. We extensively discussed microservice patterns, service modules (e.g., Bus Service, Route Service), choice of databases, and normalizations. Out of the 6 main features, I was able to cover 2-3 in detail. The interviewer asked many deep questions about my design choices. The session lasted about an hour, with only about 10 minutes dedicated to Leadership Principles. The feedback for this round was also 'Inclined'.
Round 3: HLD with Hiring Manager
My third round was a High-Level Design discussion with a Hiring Manager from India (16 YOE). After a quick introduction lasting more than 5 minutes, a significant portion of the round (35-40 minutes) was spent discussing my work, projects, and role. For the problem, I was asked to design a Book Recommendation System. I was genuinely surprised by this, as I had prepared for over 30 HLD questions and deliberately skipped this one, assuming it would involve Machine Learning, which isn't my primary expertise as a backend engineer. My mind went blank, even though I knew how Spotify's song recommendation system works and could have theoretically mapped that knowledge. I still managed to propose something, but I wasn't satisfied with my answer. As expected, the feedback was 'Not Inclined'.
Round 4: Bar Raiser
The final round was with a Bar Raiser from India (6 YOE), who was very experienced in competitive programming, friendly, and professional. After a 10-12 minute introduction and process explanation, we moved to Leadership Principle questions. The problem was a variant of the Longest Common Subsequence (LCS). The initial problem description was vague, so I couldn't identify it immediately. I attempted a basic greedy approach, but I knew it wouldn't pass all test cases, which the interviewer confirmed. He then hinted at a brute-force approach. I identified it as a dynamic programming problem and proposed a solution. He agreed it would work but mentioned it wasn't the exact intended solution. After further hints and discussion, I refined my approach and eventually solved it. Overall, I needed about 3 hints to arrive at the solution. The feedback for this round was 'Not Inclined'.
Ultimately, I received a rejection email.
Interview Questions (4)
Design a Bus Ticket Booking System. I was given 6 specific requirements to fulfill for the design, including aspects related to service design and handling concurrency during bookings. The interviewer's focus was on designing the API endpoints and the database schema, which necessitated a design approach integrating database principles with microservices. We extensively discussed microservice patterns, the breakdown into service modules (e.g., Bus Service, Route Service), choice of databases (SQL/NoSQL), and database normalizations.
Design a High-Level architecture for a Book Recommendation System. This problem required considering how to recommend books to users based on various factors. My challenge was that I had not prepared for ML-heavy system designs, assuming it wouldn't be asked for a pure backend role.
A variant of the Longest Common Subsequence (LCS) problem. The problem description was initially vague, making it difficult to immediately identify. I was expected to find the optimal solution to this sequence-based problem, which typically involves dynamic programming.
Preparation Tips
My preparation involved solving many DSA problems and practicing for over 30 HLD questions. I also made sure to behave professionally with the recruiter, as they are a key part of the process. I believe it's critical to never attempt to cheat using AI models or any other means, as interviewers are very skilled at detecting such attempts and will reject you instantly if suspicious. For LLD and HLD rounds, I would advise practicing on Amazon's Bluscape platform for design and modeling, despite its challenges. Most importantly, I realized that solving the problem is no longer the sole parameter; one must dig deeper for knowledge and understand the underlying concepts thoroughly.