First I was given a code assignment which is an algorithmic task. They claim it should be done in 2-3 hours and that probably be the case of you know exactly what to do from the start and don't care that much about code quality and tests (which they ask you to do your best at). The assignment definition is poorly written, has conceptual issues, missing information, and overall you can tell it's been given by someone who has no experience in coding tasks. For example, it doesn't become clear, but they're either really looking for the optimal solution, which would require additional dynamic programming algorithms (on top of the main algorithm calculating order processing times) searching for the real optimal solution (which doesn't fit into the 2-3 hours timeframe) or they don't realize that the solution that can be implemented in that timeframe is, in fact, not optimal.
I sent the coding task solution (with a lot of explanation on exactly what assumptions I have made, why, and how I can correct them on request). In the end, they didn't even look at it.
In the mean time, I was invited to a prescreening 30 minute interview. A significant portion of the interview was about the interviewer and his academic achievements - ok, 2 PhDs is very impressive. I got rejected at that step because I only have a few computer vision projects behind my back and none of them are related to object tracking - that is fine, they're are a startup and are looking for knowers and not learners. I also don't have enough experience with OpenCV. What surprised me the most, though, is that in their feedback they stated that I don't know model hyperparameters. How did you understand that given we didn't even discuss the projects that I've worked on nor did we we talk about ML concepts, network architectures, etc.. in that 30 minutes ? In what world could I've done any CV projects and not have a fundamental understanding of hyperparameters?
Anyway, I guess they're new at this.