Interview Stages

Round 1: Phone Screen

• Process: Started with a 30-minute discussion about my projects.
• Problem: Given coordinates representing stations, output all stations on the street with the most stations. Streets are defined by horizontal, vertical, and diagonal lines.
• Focus: Data structure and logic for counting stations.

Round 2: Virtual Onsite (VO)

ML System Design:

• Process: Began with a 30-minute discussion about my projects, followed by system design on fraud detection.
• Topics: Data processing, model comparison (decision tree vs XGBoost vs NN), metrics selection.
• Focus: Handling time-sensitive fraud events, large-scale data distribution, model quantization, and connecting models to users via agents.

Behavioral Questions:

• Process: Focused on my projects and experiences.
• Topics: Contributions to the community, ownership, and valuable feedback received.
• Focus: Communication and collaboration skills.

ML Coding:

• Process: Implemented logistic regression and binary cross-entropy.
• Problem: Derivation of sigmoid function, calculation of confusion matrix.
• Focus: Understanding of ML fundamentals and problem-solving under pressure.

Coding:

• Process: Solved a problem using two pointers and discussed divide and conquer for multiple datasets.
• Focus: Debugging skills and scalability considerations.

Overall: Each round included questions about production challenges, which I found difficult to answer.