\cleardoublepage \part{Data, Models and DevSecMLOps} \setbookpart{Data, Models and DevSecMLOps}

Part II: Data, Models and DevSecMLOps

Part II turns to the technical substance of the platform: data, the lifecycle of models, prompts and agents, the security loop through it, and the patterns that connect probabilistic services to deterministic systems of record.

The chapters share one stance: every element of the platform is a codified, version-controlled artefact with explicit boundaries. That stance lets the security controls of Part IV and the operational disciplines of Part V attach to the platform rather than be bolted on, and it lets the governance controls of AI Governance consume the platform's records as evidence.

Chapter 6 covers data architecture and Chapter 7 the model lifecycle as a pipeline. Chapters 8 and 9 build the DevSecMLOps loop and examine its weak points, from poisoned training data to over-privileged pipeline identities. Chapter 10 addresses foundation models and retrieval, Chapter 11 agentic systems with bounded autonomy, and Chapter 12 integration patterns.

What you will learn in this part:

• Data products, lineage, feature and vector stores, and synthetic data for AI
• Registries, promotion gates and rollback, and how LLMOps differs
• A DevSecMLOps loop in which security is a property of every stage
• Retrieval-augmented generation, grounding and prompt management
• Agentic systems with auditable tool use and human approval gates
• Integration of probabilistic services into deterministic enterprise landscapes

This is the book's most technical part; for the underlying automation practices it refers to Architecture as Code.