Three-volume edition
AI Engineering
Architektur, Governance und Betrieb produktiver KI-Systeme
Sebahattin Özata · Diplom-Informatiker (M.Sc. equiv.)
Three volumes, 38 chapters, and an appendix volume with mathematics, exercises and full solutions. The work is in preparation.
- Structure
- 3 volumes
- Chapters
- 38 + Appendix A
- Pages
- 1,649
- Diagrams
- 260
- Tables
- 200
- Language
- German
Engineering over hype
Why AI Engineering?
Production AI systems rarely fail on the choice of model. They fail on missing architecture, unclear data pipelines, weak governance and operational blindness.
A production AI system is more than a model. It needs architecture, retrieval, orchestration, governance, security and observability — designed precisely, not assembled by accident.
The work treats these layers as one subject: theory and mathematics connected to architecture and production engineering.
Subject area
What the work covers
- Foundations & Mathematics
- Models & Embeddings
- Data & Retrieval (RAG)
- Agentic Systems
- Architecture
- Governance
- Security
- Observability
- Operations
Three-volume edition
Three volumes. One work.
Volumes 1 and 2 form the main work with 38 chapters. Volume 3 accompanies both expository volumes with mathematics, exercises and complete solutions.
Three volumes. One work.
How the three volumes relate to one another — exposition, operations, and the mathematical underpinning.
- Volume 1
Foundations and System Components
Chapters 1–22Foundations, models, data, retrieval and agentic systems.739 pagesMore - Volume 2
Governance, Security and Observability
Chapters 23–38Architecture, governance, security, observability and operations.653 pagesMore - Volume 3
Mathematics, Exercises and Solutions
Appendix A + solutions to chapters 1–38Companion to Volumes 1 and 2: Appendix A, exercises and complete solutions to all 38 chapters.257 pagesMore
Structure
How the work is organised
The 38 chapters run across two expository volumes. Volume 1 moves from the foundations to the components of a system; Volume 2 from architecture to running operations. The break between chapters 22 and 23 separates what a system is made of from how it is designed, secured and operated.
Volume 3 sits across both. It holds Appendix A along with exercises and complete solutions to all 38 chapters, carrying the mathematical and practical underpinning of the other two. Readers who want to follow the derivations or test what they have read work through Volume 3 alongside the others, not afterwards.
The split keeps theory and practice separately readable without detaching them: the exposition stays arguable throughout, and the mathematics sits in full where it can be checked.
- Volume 1Foundations and System ComponentsVolume 1 is the first of two expository volumes and covers chapters 1 to 22.Chapters 1–22739 pages
- Volume 2Governance, Security and ObservabilityVolume 2 is the second expository volume and continues the series with chapters 23 to 38.Chapters 23–38653 pages
- Volume 3Mathematics, Exercises and SolutionsVolume 3 is the supplementary volume: Appendix A along with exercises and complete solutions to all 38 chapters of Volumes 1 and 2.Appendix A + solutions to chapters 1–38257 pages
Readership
Who the work is written for
Software and AI engineers
Building and operating production systems, not just prototypes.
Architects and engineering leads
Deciding on the structure, governance and operation of such systems.
Technical decision-makers
Assessing the adoption of AI on technical grounds.
Students and researchers
Working through AI engineering as a discipline in its own right.
Sample chapters
Six excerpts to read
Selected passages from both expository volumes — from retrieval and agents through prompt injection and testing to the question of how trust in an AI system is calibrated.
6 excerpts · chapters 14, 16, 18, 27, 33, 35
Go to the samplesAuthor
Sebahattin Özata
Diplom-Informatiker (M.Sc. equiv.)
Sebahattin Özata is a principal engineer and software architect with decades of experience in distributed systems, security architecture and AI platforms.
An English edition of the three-volume series AI Engineering is planned.
Questions
Common questions
How is the work divided?
Volume 1 covers chapters 1 to 22, Volume 2 covers chapters 23 to 38. Volume 3 contains Appendix A along with exercises and complete solutions to all 38 chapters.
Will there be an English edition?
An English edition is planned. No date has been set.
When does the work appear?
The work is in preparation. Publication date and availability will be posted on this page once they are settled.
Status
In preparation
Details on publication and availability will appear here once settled. Enquiries from libraries, technical press and about rights are welcome in the meantime.
Get in touch