Very poor and unprofessional. In the end, I angrily stated that I would write a post about the interview process, tag them to ask about the questions they posed, share my results on their images(my results were quite decent), and request them to share their current results. However, I later felt that it wouldn't be morally correct to escalate the situation to a personal level or disclose every detail of our private conversations. However, I would like to share my interview experience. So, I got a call. I only committed for a short term (2 months). I was asked a coding question(I shared below), I completed it shared it along with the results on the images for the project(not even project it was a milestone) that they were hiring for. And the results were decent and I asked if it works and the last message was let's discuss in the meeting. However, the meeting where we were supposed to discuss about the project/results turned to an interview. They started with questions like handling class imbalance(I answered data augmentation/weighted loss ), steps I would take to deploy model on edge(quantization/pruning)(again I feel questions not relevant to the project, I think it should have been more on object detection and the project like the benefits of ResNet or FPN), but I became particularly frustrated when the interviewer became overly obsessed with unnecessary calculation parts for quantization. I'm unsure what he was trying to prove, but I doubt he intended to share that quantization could lead to a 25-40% drop in accuracy. I simply asked him to provide an example of a well-behaved classification model where the accuracy dropped more than 25%, and how this question was relevant to the role. I explained that it doesn't solely depend on weights, and if logits still rank the same, you can still achieve good accuracy. The context is important for these questions. You can check Rasa, Nvidia, and other blogs that have quantized BERT without significant accuracy loss using the latest techniques. If you want to ask advanced questions, you should be prepared to answer follow-up questions. I believe it's best to ask questions about things you've personally worked on. I don't think the interviewer had personally used the latest quantization techniques or was knowledgeable about object detection models (the project he was leading). He seemed unwilling to ask questions about these topics or share his results. So, my point is since you are so obsessed with "perfection", first be clear if you are hiring for "machine learning engineer" or "machine learning researcher" and edit the jd. Second, I had already told them i am into nlp and rag(applied ml) for the last 1,5 years, if they wanted a researcher they should not have even tried to proceed with me(remember, I did not apply for this position). Again, do that in your own work then what you preach. I asked him if he has developed a new architecture or what changes he made to yolo(even the basic ones), he again did not want to answer. This is not a general feedback. I know there must be some good people/managers in the company. So, if you are applying for other roles, don't use this to decide. But for machine learning engineer role, I would not recommend it. Even if I had ended the interview peacefully and they had offered me this role, I would have rejected it. I don't like control freaks. For the machine learning engineer role, I wouldn't recommend it.