Your question is Fraud Detection ML Pipeline. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you architect a machine learning pipeline for fraud detection that accounts for model drift and retraining costs?