Describe a scenario where a project wasn't going to work, what did you do?
Machine Learning Research Interview Questions
127 machine learning research interview questions shared by candidates
Difference between generative and discriminative classifiers
A long and in detailed discusssion about my research. Recruiter was very well prepared, read my papers and asked about some details
My experience with PhysicsX was unfortunately very disappointing and frustrating. Despite being informed of a structured interview process consisting of four rounds, only two technical interviews actually took place. The first round felt more like a formality (easy leetcode problems on Coderbyte), while the second involved a technical assessment centered around 3D datasets relevant to the day-to-day work at PhysicsX. It's worth noting that machine learning (ML) hadn't even entered the discussion at this point. The promised third round, which was supposed to delve into PyTorch and ML optimization, never occurred. Instead, I received a rejection, citing the need for stronger experience in PyTorch and optimization. What's particularly disheartening is that these skills were never even evaluated. This experience not only wasted my time but also left me feeling undervalued as a candidate. I would advise others to carefully weigh the potential time investment before considering opportunities with PhysicsX.
Questions around linear regression with complex noise, ML take-home assignment.
Some questions are based on the projects listed on my resume.
Median Leetcode question, meaning that they don't have a specific expectation of skill set that the candidate should possess, other than logic and general algorithm on Leetcode.
Don't wanna give away anything specific due to confidentiality but the take home and phone screen were what's been mentioned by others. The three parts of the onsite were very fair and they are by no means trying to burn through applicants like other posts have mentioned. If you know what you've put on your resume and your own research well and can code basic ML pipelines you'll be fine.
Complete one of the projects sent via email.
Not selected for the interview.
Viewing 41 - 50 interview questions