6,49 €
Nouveauté
Architecting Production-Ready Gen AI and Agentic AI Systems
Par :Formats :
Disponible dans votre compte client Decitre ou Furet du Nord dès validation de votre commande. Le format ePub est :
- Compatible avec une lecture sur My Vivlio (smartphone, tablette, ordinateur)
- Compatible avec une lecture sur liseuses Vivlio
- Pour les liseuses autres que Vivlio, vous devez utiliser le logiciel Adobe Digital Edition. Non compatible avec la lecture sur les liseuses Kindle, Remarkable et Sony
, qui est-ce ?Notre partenaire de plateforme de lecture numérique où vous retrouverez l'ensemble de vos ebooks gratuitement
Pour en savoir plus sur nos ebooks, consultez notre aide en ligne ici
C'est si simple ! Lisez votre ebook avec l'app Vivlio sur votre tablette, mobile ou ordinateur :
- FormatePub
- ISBN8235354739
- EAN9798235354739
- Date de parution21/07/2026
- Protection num.pas de protection
- Infos supplémentairesepub
- ÉditeurIoakim Ioakim
Résumé
A successful AI demonstration proves that a model can respond. A production-ready AI system must prove much more. It must operate within defined authority, retrieve the right enterprise knowledge, protect sensitive data, survive model and provider changes, control cost, expose its behaviour through telemetry, and produce evidence that engineering, risk and governance teams can examine. Architecting Production-Ready Gen AI and Agentic AI Systems is a practitioner-focused guide to designing generative AI, retrieval-augmented generation and agentic systems for real enterprise environments.
The book moves beyond isolated prompts and model APIs to examine the architecture surrounding the model: AI gateways, orchestration, retrieval pipelines, memory, tool contracts, evaluation systems, security controls, observability, FinOps, human approval and governance evidence. Using the recurring NovaCred case study, the book demonstrates how a regulated organisation can progress from policy-grounded generation to controlled agentic workflows without allowing autonomy to advance faster than its operational and governance maturity.
Readers will learn how to:. Distinguish AI-added, AI-first, generative, agentic and hybrid systems. Assess organisational readiness and system maturity before increasing autonomy. Design a canonical enterprise architecture for production AI. Build reliable retrieval and knowledge-grounding pipelines. Define tool permissions, action tiers, approval gates and rollback paths. Architect single-agent and multi-agent systems.
Treat prompt engineering as a governed engineering discipline. Create evaluation suites, release thresholds and regression controls. Defend against prompt injection, data leakage and excessive agency. Measure model, retrieval, tool and workflow behaviour in production. Control token consumption, model-routing costs and agent execution budgets. Convert governance requirements into runtime controls and auditable evidenceThis book is written for enterprise architects, AI engineers, platform leaders, technology executives, product owners, data professionals, security teams, risk officers and governance practitioners responsible for moving AI systems from experimentation into dependable production.
The central principle is straightforward: the model is only one component. Production readiness is a property of the entire system.
The book moves beyond isolated prompts and model APIs to examine the architecture surrounding the model: AI gateways, orchestration, retrieval pipelines, memory, tool contracts, evaluation systems, security controls, observability, FinOps, human approval and governance evidence. Using the recurring NovaCred case study, the book demonstrates how a regulated organisation can progress from policy-grounded generation to controlled agentic workflows without allowing autonomy to advance faster than its operational and governance maturity.
Readers will learn how to:. Distinguish AI-added, AI-first, generative, agentic and hybrid systems. Assess organisational readiness and system maturity before increasing autonomy. Design a canonical enterprise architecture for production AI. Build reliable retrieval and knowledge-grounding pipelines. Define tool permissions, action tiers, approval gates and rollback paths. Architect single-agent and multi-agent systems.
Treat prompt engineering as a governed engineering discipline. Create evaluation suites, release thresholds and regression controls. Defend against prompt injection, data leakage and excessive agency. Measure model, retrieval, tool and workflow behaviour in production. Control token consumption, model-routing costs and agent execution budgets. Convert governance requirements into runtime controls and auditable evidenceThis book is written for enterprise architects, AI engineers, platform leaders, technology executives, product owners, data professionals, security teams, risk officers and governance practitioners responsible for moving AI systems from experimentation into dependable production.
The central principle is straightforward: the model is only one component. Production readiness is a property of the entire system.
