```\nStatus: CS PhD student, Tier-1 college\nPosition: Research Scientist Intern, Amazon (Computer Vision)\nLocation: Seattle\n```\n\nI did not apply for the position but recruiter emailed me saying someone from team reviewed my profile (my website and publications) and would like set up interviews. \nI had total of 3 interviews scheduled, each on 3 consecutive days. The scheduling process was very quick (within 2 days, not time to prepare) \nAll interviewers were senior scientists at Amazon (5+ yrs exp)\n\n```\n1st interview (Phone, 60 mins)\n```\n1. Interviewer asked about ongoing projects, got interested in a GAN-project I was working on. Asked me to implement a small logic from my idea (10-15 loc) Some specific details about training GAN model (since it is hard to stabilize its training) \n1. Coding Q1: Find length of longest sequence of same consecutive numbers in array (easy, solved in 2 mins)\n1. Coding Q2: Implement image-based caching system for fast retrieval. Similar to [{this}](https://leetcode.com/problems/lru-cache/) but focus on image processing and storing part (solved)\n1. 2 Questions on overfitting and bias corrections in ML models (lot of followups)\n1. 2 questions based on Amazon leadership principles from past experience\n\n```\n2nd interview (Phone, 60 mins)\n```\n1. Started with introduction, came to asking the theory and math of ResNets. And also implementation of it loose syntax. \n1. Coding Q1: Video based recommendation system (like Netfilx) but with only one feature. (solution was to identify it as collaborative filtering problem & implement it)\n1. 2 questions based on Amazon leadership principles from past experience\n\n```\n3rd interview (Phone, 30 mins)\n```\n1. Open ended questions on my research projects, what steps I considered before coming with solution etc.\n1. 2 questions based on project I was going to work on, how I would approach them (can\'t share details)\n\n\nGot offer after 2 days\n\n```\nThoughts\n```\n* I think questions were well thought and I could only answer because of my experience implementing deep learning models, especially follow-ups were pretty real. They expect you to have proficiency of leetcode-medium but in addition fair understanding of which ML/DL algorithms to use when and its limitations. Also the project was relevant to my PhD projects so they read about my work and already knew my proficiency on some topics.\n

\nStatus: CS PhD student, Tier-1 college\nPosition: Research Scientist Intern, Amazon (Computer Vision)\nLocation: Seattle\n\n\nI did not apply for the position