This week covered "how to release trained models to production." Inference option selection (Day 1), real-time endpoint operations (Day 2), inference optimization (Day 3), deployment strategies (Day 4) look separate but form one decision flow: which inference mode → which endpoint config → how to optimize → how to deploy safely. Today we compress this flow into one page and organize keywords that split answers on the exam.