I applied online. The process took 1 day. I interviewed at Graphcore (London, England) in Aug 2025
Interview
I received an email about a month ago with a link to an online test. Please attempt it after practicing MCQs or revising linear algebra, data structures, and algorithms, as the test is quite difficult. After that, I had an initial call with the lead, who explained the interview process. He mentioned that there would be four to five rounds in total.
The first stage was the live coding exam, which lasted an hour. The next stage would have been a paper discussion, involving reviewing and discussing a research topic. Unfortunately, I was only able to progress up to the coding exam, after which I was rejected.
Interview questions [1]
Question 1
Online Test: Take the HackerRank online test. Make sure to review Big O notation beforehand. You’ll have 20 minutes to answer 20 difficult MCQs.
Live Coding Exam: The type of question you get is unpredictable, it really depends on luck. However, it’s important to make the session interactive. Try not to stay silent while coding; instead, explain your thought process as you go.
4 stages.
The first one is the test. 20 questions, mostly underground maths and coding.
The 2d interview is simple, just a call with someone in the company. More like get to know each other. They are keen on someone with robotics experience. I was rejected after that stage.
They are very fast with decisions. I think I only waited a week or less after the application for the first stage and a couple of days between each stage.
Interview questions [1]
Question 1
The test was in hackerank. Few were on matrices (find a rank, dimension etc). One-two were someone integral or derivative (not school level). Others were on tensor code or some coding questions, pretty simple. And a couple were on O(n), algorithms.
The interview was 30 mins, including some technical questions, and a coding excerise. They asked me 2-3 questions first, then we started the coding excerise. The coding excerise was not difficult.
Interview questions [1]
Question 1
If I wanted to evaluate an optimizer for deep learning methods (e.g. Adam), how would I approach it?
