Your question is Automated Retraining 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 an automated retraining pipeline for a live model at Zillow?
Focus on how you would detect drift, trigger retraining, validate new models, and safely promote them to production. Assume the model is user-facing and business-critical, so downtime, skew, and bad rollouts matter.
