I applied through an employee referral. I interviewed at Google DeepMind in Oct 2021
Interview
I had a referral.
Did an initial HR screening => Technical quizz ==> Coding interview ==> Interview with a researcher. At this point I got a rejection email saying that the current research priorities aren't a good fit for my research interests and strengths, but that they will keep my profile in mind for the future. (I think I did well in all those parts, so I am giving 50% to the possibility that the second part was actually true. )
The interview with a researcher was a bit confusing, because they didn't very clearly communicate the purpose of it. I mostly got asked trivial or vague questions which made it seem like they are just testing whether I understand the topic. But they might have been trying to assess my fit with the group's research interests. But I don't think the confusion was deliberate.
If I was successful at the interview, next steps would have been one hour candidate talk => ~5 more interviews with researchers => "people and culture" interview.
At each point, they give you tips on how to prepare for the next part. The tips for the technical quizz & coding interview were imo sufficient to adequately prepare.
Interview questions [3]
Question 1
Initial HR screening:
- How are you today?
- How was your weekend?
- Why DeepMind?
- AGI --- do you think we will have AGI?
- Tell me about your research.
- Asked more details about collaboration with X.
- What would you like to work on in DeepMind?
- What style of work works best for you?
- When would you ideally start?
- Are you applying to any other places within Industry? Academia?
- Any things in terms of timelines? Deadlines we should be aware of?
Coding interview
- Part 1 (theory and practice, general chat)
- All questions were very open-ended.
- Difference between [two of your languages of choice].
- Practical implications of the differences?
- When you use variables in C++ vs Python, what are the types? (Hinted at python variables not having a fixed type.)
- Memory leaks, memory management. Python vs C++.
- Private variables in C++ and Python. So does this mean that you can't use Python to write secure software?
- Object oriented programming. What are the advantages? When can it be good?
- How would you implement Python dictionary if Python interpreter didn't exist?
- Explained implementation using hash table.
- Q: Another way of implementing it?
- All complexity results from the start of the century hold. But some things have changed in our software and hardware. How does it affect things?
- Part 2:
- Showed me a google slides presentation describing a geometric algorithm for calculating the value of pi.
- Task: implement it (took longer in the first part, so only had 15 minutes).
- (Roughly: Approximate arc by piecewise-linear function, calculate length of a segment, multiply by number of segments.)
- At the end, asked which extra steps would I do if this was to be usable by other people.
Interview with a researcher:
- What are you reasons for wanting to join team X at DeepMind?
- Which papers coming out of team X did you find interesting?
- Specific questions about the area in which I do research.
- Space for my own questions.
I applied online. I interviewed at Google DeepMind
Interview
There will be one coding round and one ML round, and each round will last one hour. The ML round mainly consists of oral questions about core machine learning concepts.
Interview questions [1]
Question 1
The ML round mainly consists of oral questions about core machine learning concepts.
I applied through an employee referral. I interviewed at Google DeepMind (Mountain View, CA) in May 2026
Interview
The interview process was quite smooth. I had two research and two coding interviews. Research interviewers asked what I would like to do at GDM if I join. One of the research interviews was quite confrontational, but respectful. All the interviewers were kind.
7 -9 rounds depending on candidate. The rounds involved 1-3 scientists taking test on leetcode style questions, AI fundamentals, Transformers, Finetuning, Training approach of large models like Gemini, Deepseek etc.
Interview questions [3]
Question 1
Straight in the interview I was asked a coding question leetcode style.
Research Scientist applicants have rated the interview process at Google DeepMind with 3.3 out of 5 (where 5 is the highest level of difficulty) and assessed their interview experience as 43% positive. To compare, the company-average is 63.6% positive. This is according to Glassdoor user ratings.
Candidates applying for Research Scientist roles take an average of 45 days to get hired, when considering 7 user submitted interviews for this role. To compare, the hiring process at Google DeepMind overall takes an average of 40 days.
Common stages of the interview process at Google DeepMind as a Research Scientist according to 7 Glassdoor interviews include:
Phone interview: 38%
One on one interview: 25%
Presentation: 25%
Skills test: 13%
Here are the most commonly searched roles for interview reports -