Graphcore interview questions

After getting in touch with the recruiter, I was sent a Hackerrank test to test C++ skills. It's not the easiest test, but it's doable. Afterwards I was invited for an interview with the manager of the team and afterwards an interview with two software engineers from the team.

I contacted a person from GraphCore on LinkedIn to ask about projects and whether they have projects related to Biology, which it seems they did. - Filled up a long form when applying online describing motivation for the role - Web interview with 2 members The interview was scheduled for 10 AM but did not send a calendar invite. Due to a mixup in TimeZones I was 5 minutes late but we start the interview. I was asked very specific questions about how numbers are represented in computers and how deep learning works. My interviewers asked about whether the data or the model is larger, and I replied that it depends on the batch size, implying that "data" could be 1 image or all of your images. The interviewers took it to mean that I thought that with larger batch size the model would be larger. The question was a bit ill-posed and I personally work with models that are Gigabytes in size. The "data" is also ambiguous as data may be transformed to fit in a model, and that transformation (for example voxelisation from coordinates) can increase the size of your "data". Your model may still be larger than your full original dataset of cartesian coordinates (essentially it's a list of int/floats vs a 3D one-hot encoded array ) I got my rejection a week later at the beginning of April 2022. I asked for feedback and the recruiter replied she could call me the next day. She did not. I email once a week to ask about this, and got no response until May 2022 (a month later). She called me 10 minutes later the scheduled time. The feedback was that my professional approach was poor and I was not "apologetic enough" for being 5 minutes late. Although I explained the mixup and that in the future a calendar event would have been better, the recruiter insisted this was the opinion of 2 interviewers. During the interview I was asked about how I found out about GraphCore and I said that I researched chip manufacturers in the EU and found out about their tech and how I find it really interesting, especially in applied machine learning contexts. I said that their technology was "freaking cool", however the recruiter said I had sworn during the interview and that made me "unprofessional". This is fair, although I explained that I did not in fact swear she continued saying how unprofessional I was for denying it. When I asked if we could skip this feedback and move to the technical feedback she said I was defensive again and being unprofessional and she did not see a reason to continue the call. This is rather ironic coming from someone that took a month to reply to an email and was 10 minutes late for a call and is criticising someone for being late. She did eventually provide feedback that I have poor/non-existing knowledge of "deep learning". I guess they are really looking for Computer Scientists or people with strong theoretical knowledge rather than applied Machine Learning engineers, which is fine, however I had made it clear in the application that I was interested in the applied projects. My advice would be to review the fundamentals as you won't be asked to solve problems but it'd be more like them going through a textbook and you having to describe the correct answer (at least for the first part of the interview). Regarding my experience, I would honestly not recommend it. A tech company that does not know how to send a calendar invite, complains about not people being "apologetic enough" for being 5 minutes late and then responds with a month of delay and 10 minutes late for a call, is simply incoherent.