Scale Machine Learning Engineer Interview Questions
& Process

The overall interview process was good but demanding. There were 7 rounds in total: a recruiter call, a technical call focusing on Object Oriented Programming, a manager call, and finally a virtual onsite with 4 back-to-back interviews. These onsite interviews covered technical aspects (OOP), behavioral questions, applied ML, and ML Fundamentals. The process itself was clearly communicated, and they used ModernLoop for scheduling and preparation. However, it was very time-consuming, with over 7 hours of interviews. The ML Fundamentals round involved reading a research paper and answering questions about it. It's important to be optimistic about LLM and AI research; my opinion that ML research wasn't fundamental seemed to be poorly received by the interviewer. Ultimately, I was denied for not performing well in the ML Fundamentals interview, which is very disappointing after investing so much time. It also makes me question working with people who can't accept different viewpoints or critically assess their own processes, especially when doing well in all other interviews.

So a recruiter reached out to me via email about a potential role. They were interested and gave me a deep learning project to work on. The project was about detecting the orientation of stars in images. You know, like taking a picture of the night sky with your phone, connecting five random stars to make a giant star, and then calculating its orientation. I had two weeks to complete this. I had the choice between a computer vision project or an NLP project, both came with their own datasets and skeleton code. They wanted to make sure I really understood Pytorch and Python.

It was a nice and smooth online assessment. The response time from the first round and HR was really fast. There was a take-home assignment on HackerRank with two questions (the same problem). It was pretty difficult and took time to resolve, and also involved learning itself.