Скачать или смотреть Part II — Chapter 1 : Design Principles — AI Workflow

a comprehensive overview of the Machine Learning (ML) lifecycle as a systematic framework essential for building production-ready artificial intelligence systems. This framework is detailed through six core stages: Problem Definition, Data Collection & Preparation, Model Development & Training, Evaluation & Validation, Deployment & Integration, and Monitoring & Maintenance. A central theme is the fundamental difference between ML workflows and traditional software development, highlighting that ML requires iterative experimentation, continuous feedback loops between stages, and robust strategies to handle data drift and dynamic performance degradation. The text uses a diabetic retinopathy (DR) screening system case study to illustrate how real-world constraints, such as edge device limitations and regulatory compliance, drive interconnected decisions across the entire lifecycle, demonstrating the necessity of a systems thinking approach.