Amazon — Machine Learning Engineer ❌ Failed

Level: Senior-Level

Round: Phone Screen · Type: Coding · Difficulty: 6/10 · Duration: 60 min · Interviewer: Neutral

Topics: Data Structures, Algorithms, System Design

Location: Seattle, WA, US

Interview date: 2026-01-15

Got offer: False

Summary

I had a phone screen interview for the Audible team at Amazon. The first 30 minutes were spent discussing my resume, and the remaining 30 minutes were dedicated to a coding question.

Details

The coding question was:

`

we want to detect "abusive books" in a stream of user read event.

event looks like this: <user_id, book_id, current_pos_in_sec, book_duration_sec>

an event is abusive if:

(1) user reached the last 5% of book_duration_sec (e.g. only consider​ those who almost complete the content)

(2) when (1) happens, same user hasn't completed 10% of book_duration_sec (e.g. abusive books is boring, so people skip ahead)

a book is considered "abusive" if abusive_event_ratio > 20%

given stream of event:

readBook(user_id, book_id, current_pos_sec, book_duration_sec>

write a solution to detect "abusive books"

`

My general idea was to use a global cache to store <(user_id, book_id): progress_sec>, <book_id: (abusive_event_cnt, total_event_cnt)>. I discussed top k, but I was told not to consider it. I asked if events could be repeated, and I was told to assume they were not, since there wasn't much time. I discussed atomic counters and wrote a pseudo function, but I didn't implement a lock or atomic incrementer.