Runway Machine Learning Engineer Interview Experience — Two Phone Screens, Rejected in ML System Design

Sharing an interview experience for a lesser-known company. It's a video generation startup that seems to be doing pretty well lately. I mass-applied to one position, but the recruiter contacted me about a different role instead. There were two phone screens total, both 45 minutes.

Round 1 was coding: implement an ngram class with fit and generate methods. fit reads a file and counts ngrams, and generate does simple text generation based on the ngram probabilities. The follow-up was to find the optimal n — I didn't have time to write code for that, so we just briefly discussed the approach.

Round 2 was ML system design: design a video search system, covering both the model and the infra. The question itself wasn't hard, but the interviewer kept interrupting and getting stuck on some modeling details. I tried explaining it in various ways, but the interviewer still didn't seem to get it, and we ran out of time before I could finish. Looking into the interviewer's background afterward, it seemed like they were from ML infra and maybe didn't know modeling that well. I got cut after this round.