Round 1: Machine Learning Discussion

In my previous organization, I worked on a BERT-based Multi-Task Learning (MTL) model. Most of the discussion during this round revolved around BERT and MTL architectures. Below are some of the questions I was asked:

• Explain the architecture of the BERT model.
• What are positional encodings, why are they needed, and what types of positional encodings are you aware of?
• What is the objective function used during BERT pre-training?
• What is the Next Sentence Prediction (NSP) task, and how would you construct a dataset for training this task?
• How is RoBERTa different from BERT?
• How does an MTL model function? What objective function would you use in an MTL setup? Can you write the mathematical equations?
• How would you approach hyperparameter tuning in an MTL-based model?

Round 2: DSA Coding

There were two coding questions in this round:

• You are a data scientist at an AI research lab. Your team is developing a win predictor model for the game of Snakes and Ladders. For this, you are tasked with simulating realistic gameplay data. Write code to generate a random Snakes and Ladders board and simulate a two-player game on it. The output should include the sequence of game events.
• Given the root of a binary tree and an integer targetSum, return the number of paths where the sum of the node values along the path equals targetSum. The path does not have to start or end at the root or a leaf, but it must move downward (i.e., from parent to child).
  • Example 1:
    Input: root = [10,5,-3,3,2,null,11,3,-2,null,1], targetSum = 8
    Output: 3
    Explanation: The valid paths that sum to 8 are highlighted.
  • Example 2:
    Input: root = [5,4,8,11,null,13,4,7,2,null,null,5,1], targetSum = 22
    Output: 3

Round 3: System Design Round

Design a system that verifies whether all required compliance steps were followed during a KYC (Know Your Customer) process by a company agent.

• Check if the agent verbally mentioned required compliance statements, such as notifying the customer that the call is being recorded.
• Detect if multiple individuals are present in the video background.
• Validate whether the PAN card is genuine.

Hiring Manager Round: Behavioral

• Describe the most impactful project you have worked on.
• Have you ever had a conflict with a team member? How did you handle it?
• Have you collaborated with non-technical stakeholders? How would you explain a technical concept to a non-technical audience?
• A few more questions were asked, similar to Amazon’s behavioral interview format.