Python in data/ml rarely goes into that scale. It is used for Training. Several thousands line per project at most, it is tractable and I don't think machine learning models really can be refactored or debugged like a web application.