Carter The Enterprise MLOps Playbook

How to Deploy, Govern, and Scale Machine Learning Systems in the Real World
1. Auflage 2026
ISBN: 979-8-23404890-5
Verlag: PublishDrive
Format: EPUB
Kopierschutz: 0 - No protection

E-Book, Englisch, 458 Seiten

The Enterprise MLOps Playbook is the operating manual for engineering and leadership teams responsible for moving machine learning from proof of concept into reliable, governed, scaled production. Written for ML engineers, platform leads, data science managers, and the directors and executives who fund and govern enterprise AI programs, the book treats MLOps as a connected system of architectural, engineering, governance, and organizational disciplines rather than a tooling problem.
Across fourteen chapters organized around three imperatives-Deploy, Govern, and Scale-the book follows Priya Raman, Senior Manager of ML Engineering at Helios Industrial, a Fortune 500 manufacturer with hybrid infrastructure and a portfolio of ML use cases that range from predictive maintenance and demand forecasting to computer vision on the production line. Through Priya's investigations, escalations, and architectural decisions, the chapters develop the patterns that close the deployment gap, control regulatory and reputational risk, and scale ML platforms from a handful of models to hundreds without proportional cost.
The book opens with the production imperative and the structural causes of the deployment gap: the orphaned pilot, the handoff failure, the CI/CD gap, and the governance vacuum. It then advances through platform architecture, data engineering, model development, deployment engineering, and ML observability to complete the core production foundation. From there, the book extends into the operational categories that have emerged as first-class MLOps concerns: LLMOps for large language model deployment, multi-modal systems with temporal alignment requirements, and agentic operations where autonomous systems take consequential actions inside bounded, monitored authority.
Governance is treated as engineering, not paperwork. Chapters cover fairness testing, bias measurement, explainability infrastructure, audit trails, model cards, and the regulatory frameworks-NIST AI RMF, ISO 42001, the EU AI Act, and sector-specific regimes-that define defensible production. The scaling chapters address platform standardization, multi-tenancy, cost engineering, and the operating models that allow centralized platform teams to support distributed ML product teams without becoming bottlenecks.
Each chapter pairs a Geeks Only section-architecture diagrams, pseudocode, and implementation patterns for engineers building the platform-with a Manager's Decision Guide that frames the questions, signals, and investment trade-offs leaders must navigate. Case studies, including the Meridian Capital Partners deployment failure, ground the patterns in real consequences. The result is a book that engineers and managers can read together: a shared operating playbook for the discipline that determines whether enterprise ML investment produces durable capability or expensive experiments.

Autoren/Hrsg.

Carter, Jazper