I applied through an employee referral. The process took 2 weeks. I interviewed at Amazon (Seattle, WA) in Dec 2024
Interview
I got two rounds of interviews. Both round include very similar structures: 30min research experience + 20min coding + 5min leadership principals + 5min casual chatting.
For me, the research experience part is the most critical. You can imagine you need to do a job talk without using slides. Make sure let the audiences understand what you were done is the first priority.
The coding part for both rounds are pretty easy. I believe Leetcode easy-medium levels? I don't have too much experience in practicing leetcode, so may be the estimated level is incorrect.
The behavior part is pretty hard for me. If you don't have any behavior interview experiences as me, try to find some FAQs from the internet.
Interview questions [1]
Question 1
If you will be a tech lead to develop an LLM, which architecture you will choose? (i.e., decoder-only or encoder-decoder) and why?
I was asked basic knowledge in deep learning and machine learning. Also had time of explaining my research. discussed how I can apply my research to the current project. Transformer architecture, bias-var tradeoff, use of positional encoding, long-term dependencies.
HackerRank assessment with solid, fair questions. Communication with the recruiting team was clear and professional throughout the process. I was invited to two additional interviews, one focused on research depth and the other on coding skills.
One phone screen on LeetCode-style medium coding question plus behavioral questions. One loop of three back to back interviews including one round of coding, two rounds of research plus behavioral questions.