The process started with a recruiter screen, followed by a Python/PyTorch coding interview. The use of an AI tool was allowed during the coding interview. The problem involved classification and discussion around small datasets, preprocessing, data balancing, model choice, and evaluation metrics. The interviewer was very nice, collaborative, and clear. The second round was a system design interview. The exercise involved a medical imaging problem. This round felt much more like a technical deep dive than a broad system design interview. The interview style in the second round felt less collaborative and more judgmental.