Goldman sachs || Associate || Rejected

I have over three years of experience at a product-based finance MNC. Recently, a recruiter reached out to me for an Associate role at Goldman Sachs, and I went through a total of four rounds.

OA Round

The Online Assessment was conducted on Hackerrank and featured two easy to medium-level questions. I managed to solve both well within the two-hour time limit.

First Round (Coderpad)

This was a one-hour round, starting with introductions. I was then asked two coding questions, typically one medium-hard and one medium DSA question.

1. The first question was Target Sum. I initially presented a brute-force exponential solution and then optimized it using top-down DP to achieve O(n*sum) complexity. However, the interviewer was expecting a more optimized solution which I couldn't find in time, so I coded the O(n*sum) version.
2. The second question was a variation of Remove One Element to Make the Array Strictly Increasing, where instead of removing one element, we could delete one subarray. We only had time to discuss the approach, not to code it.

Superday Interviews

I attended the Superday rounds while running a fever because I only had one month left in my notice period, and I had heard that GS typically takes about a month to release an offer letter. I felt my concentration was a bit lacking due to the fever. I'm unsure if it was a bad decision or part of a larger plan, but it's a good reminder to be mindful when scheduling your Superday.

Second Round (Data Structures)

This was another one-hour round with a panel of two interviewers. After introductions, they asked two coding questions:

1. Given a map<child, parent> representing graph edges, I had to find the root of the tree with the maximum number of nodes. I proposed a solution with O(E) time and O(min(V,E)) space, but the interviewer hinted at further optimization. Due to time constraints, I coded my O(E) solution, but I mistakenly returned the maximum number of nodes instead of the root itself, which she pointed out at the end of the interview.
2. The second question was Sliding Window Maximum, which was a known problem to me. I started with a brute-force O(n*k) solution and then optimized it to an O(n) approach using a sliding window. I struggled a bit with running and debugging, but the second interviewer was friendly and gave me a couple of extra minutes, allowing me to successfully run the code.

Third Round (Software Engineering Practice)

This was also a one-hour round with two interviewers, starting with introductions from both sides. They asked the following questions:

1. Given an array, I needed to return true if there exists a triple of indices (i, j, k) such that i < j < k and nums[i] < nums[k] < nums[j], otherwise return false. I began with a brute-force O(n^3) solution, then optimized it to O(n^2). The interviewer expected further optimization, which I couldn't provide. I coded the O(n^2) solution, and although it failed initially due to a negligible check, which I pointed out, he seemed satisfied.
2. The second interviewer, who was very experienced, asked me to perform a Low-Level Design (LLD) for SnakeLadder. I wrote a few classes and also implemented the game logic for playing. He was almost satisfied but mentioned it wouldn't work in a distributed environment and was expecting the logic to be implemented in an event-driven manner, which I couldn't grasp at the time. I regret not asking for more clarification on his exact expectations.

There were two more rounds scheduled: 'Software Design and Architecture' and 'Hiring Manager Round', but I was rejected after the 'Software Engineering Practice' round, which I didn't expect.

Later, I thought about how the SnakeLadder LLD could be done in an event-driven, distributed environment. One solution that came to mind was using a choreography design pattern where events are published to Kafka topics. A service (e.g., diceService, boardService, or gameService) would pick up an event, process it, and publish the next event to another Kafka topic that the subsequent service is listening to. This could be achieved using libraries like KStream, or even a single Kafka topic by segregating different events based on Kafka headers. I couldn't think of distributing this small piece of code during the interview.

I did not receive any feedback despite trying to get it via email and calls to HR.

I hope this experience helps others in their preparations. All the best, champs!