Google DeepMind Research Engineer Interview Questions
& Process

I applied in Dec 2021 and heard back in March 2022. They gave me some study stuff for CS, Maths, Stats, and ML. Then I had a technical interview where they asked me some practical questions about integrals and statistical theorems.

I had an initial chat with a hiring manager and a recruiter. From those conversations, it sounded like it would be a great fit and that they would value me for all my experience. They described work that sounded like it would capitalize on my background while still allowing me room for growth. The recruiter acknowledged that I should be hired into a higher level role. I had one interview with a Research Scientist in London and two interviews with Software Engineers in Mountain View. The interview with the Research Scientist was interesting. They asked good questions that gauged problem solving and creativity. We had a good conversation. I felt like he was respectful, collaborative, and appreciated my background. Both of the interviews with the Software Engineers were disappointing. They asked questions appropriate for someone with less experience and were focused more on implementation rather than research. That would have been fine if they asked language-agnostic, algorithmic questions. But, the questions were obviously designed for python and were mostly testing off-hand knowledge of specific built-in text parsing functions. It was the sort of thing that anyone could have done easily in 5 minutes at their desk with the ability to look things up, but python is not my primary language and I don't do much text parsing. Both of them were very unwelcoming. Their comments made it clear that they had very rigid ideas about career paths and had a very different idea about the position for which I was being interviewed than the hiring manager. It very much felt like they were gatekeeping.

The interview process had three stages. First, a recruiter call. Second, two coding interviews and one ML fundamentals interview. Third, ML design and an interview with the Lead Researcher. I didn't do well in one of the coding interviews at the second stage. The coding interviews were Leetcode mediums, and the ML fundamentals interview covered optimization, regularization, loss functions, transformers, and practical training/inference questions.