Applied online. About two weeks later, a Snap recruiter reached out for an initial screen.

Technical screen (60 mins):

• 10 mins: Introduction, discussing background and projects.
• 10 mins: ML fundamentals.
• 10 mins: Behavioral.
• 30 mins: LeetCode medium, which the interviewer later converted to hard.

She gave the following template for the onsite:

• ML fundamentals (60 mins)
• ML Applied (60 mins)
• Informal Q&A with MLE (30 mins)
• ML System Design (60 mins)
• Coding (60 mins)
• Coding (60 mins)

Interviewers were only focusing on fixed-template questions, with no interest in my projects or discussing models.

ML Fundamentals The interviewer asked a lot of basic ML questions (metrics, unbalanced data, overfitting, optimizers, etc.).

ML Applied There was a mismatch between what the recruiter had told me this round would be about and what the interviewer was focusing on. The recruiter told me this would be about discussing trade-offs between different modeling approaches when solving a real-world problem. It was totally opposite during the interview. No modeling-related questions were asked, and the focus was on designing recommender systems in production. I had told the interviewer that I had no deployment experience in e-commerce projects, but he was only interested in discussing these.

ML System Design I was with a manager on designing a recommender system. He wouldn’t let me go into the modeling part and kept interrupting me before I answered his previous question.

Coding Two rounds of 1 hour each: LC medium & hard.