Engineering and Responsible Deep Learning

Part V

A model becomes consequential when it enters a system: deployment introduces drift, dependencies, scale, governance, and human impact.

The preceding Parts explain how deep-learning systems learn, perceive, generate, and act. This final Part asks what happens when those capabilities leave the experimental setting. Chapter 14 addresses the production gap, model compression, deployment architecture, monitoring, drift, and MLOps. Chapter 15 widens the frame to infrastructure, compute supply chains, labor, environmental cost, information ecosystems, and governance. Chapter 16 synthesizes the book’s ethical threads and examines the practitioner’s responsibility under uncertainty and competing values.

The final transition is from model performance to system stewardship. Reliability, auditability, governance, and professional judgment are treated as engineering requirements because technical systems operate within institutions and affect people.