I applied through a recruiter. The process took 2 weeks. I interviewed at Amazon (Seattle, WA) in Sep 2025
Interview
One applied scientist from Amazon interviewed me. She asked several questions from machine learning model, after that one data structure question involving time complexity questions. Later, I have asked 2 questions to her and she answered it.
Interview questions [1]
Question 1
1. What are the evaluation metrics used for LLM evaluation?
2. How is LSTM different from Transformer architecture?
3. Why is RAG architecture used instead of fine-tuning LLM models?
4. How is BERT different from GPT model?
5. What will I do in case of imbalanced dataset?
6. How is linear regression different from logistic regression? What types of loss functions are used in each of these scenarios?
7. What is the task of dropout and activation function?
Phone screen interview
Then onsite interviews: 5 interviews two on coding and three on science and a tech talk
You choose a topic for the presentation
The science interviews are: one on science depth, one on science breath and one on science applications
Interview questions [1]
Question 1
Question on optimizers; comparing gradient decent, adam optimizer etc
Interviewed with 1 phone screen, 1 coding, 2 ml design and 2 lp rounds. Most questions were non-leetcode questions more related to day to day ml implementations. The questions were very practical.
Difficult interview
Application
I interviewed at Amazon (Seattle, WA)
Interview
Applied for Amazon AGI. After first round, it will go into full round of multiple interviews. Lots of modern LLM training technic questions. There are still some behavioral questions, but less than general Amazon roles.