G-Research interview questions

did online test, combination of maths, prob, ML. then went on to triage interview. Honestly, out of all the interviews i have, i do not respect G-Research interview. I really think the interviewer is an idiot. I was expecting the interviewer would try to find out my problem solving skill and logical thinking. But none of that. He asked me about deriving eigenvalues, i showed him n equations to solve for n-polynomial equations. Then he asked me about time complexity and memory complexity of an ML model. And ways to optimise, i gave like a few dozen tricks such as mixed precision training, storing intermediate values onto disk (accelerator), replacing full attention to group attention etc. The feedback was, he was thinking about recomputing the activation. so what? My point is, both these questions are just knowledge points, and nothing practical or anything that tests your problem solving and logical thinking. I asked the interviewer if they even look at the financial data or do analysis on it. He told me he does model iteration but uses data handed over to him with features already built. Now it's important to understand the autoregressive models and attention mechanisms are computationally expensive and slow. If people really want to save memory, there're like a few dozen tricks by changing a few line of codes using frameworks. I really don't see how these questions are relevant to actually doing ML and solving problem, do they implement backpropagation themselves? No, do they use popular frameworks? Yes. Honestly, i respect all the other interviews i had with big/small companies. Not this one from G-Research. The most stupid interview i've ever had.

1) Recruiter screen 2) Discussion with lead engineer and management 3) Technical Interview split into 3 sessions. 3.1) Monitoring and Troubleshooting. Discussing how you would implement logging for example. Followed by troubleshooting a problem. No implementation. 3.2) Programming session: Write some routines in a language of your choice. Shifting numbers for example. Followed by discussion of a project that you implemented 3.3) K8S deep dive talking through how k8s works.