I applied through an employee referral. I interviewed at Amazon (Seattle, WA) in Jan 2021
Interview
The interviews are virtual due to Covid-19. So the normal call interview was more like an in person interview.
The interviewer went over my CV and asked for every project I worked on. He also asked about my PhD thesis, which was not related to the job at all.
Interview questions [1]
Question 1
Coding question (Python):
Assume you have a file containing data in the form of
data = [{"one":a1, "two":b1,...},{"one":a2, "two":b2,...},{"one":a3, "two":b3,...},...]
How could you split this data into 30% test and 70% train data?
Looking back, I'm relieved I declined the offer, despite the intense experience. The interview process felt overwhelming, starting with some tough core ML concepts before diving into the LLM fundamentals. During the technical round, I recognized a tokenization question from a PracHub session I had done just a week before. It felt like a small win in an otherwise challenging interview. Ultimately, the pressure and expectations were high, but I felt it wasn't the right fit for me.
Interview questions [2]
Question 1
LLM fundamentals: tokenization design and KL-regularized SFT
There are three rounds in total. The process begins with a coding round, followed by the main interview loop, where you will meet the team and discuss technical skills, experience, and fit.
First round is fun, second round, which is also the final round involved 5 sessions, with different focus. For some sessions, not be able to present my story completely, time was tight, and interviewers were rushing.