Mastering Agentic AI: Architectures, Design Patterns, and Production-Grade Systems

Agentic AI is easy to demonstrate and hard to ship. Mastering Agentic AIis a hands-on engineering guide that follows one Master 2 course in AI Systems Engineering, lecture by lecture and lab by lab, and turns it into five chapters you can read and run.
You will learn how to:
move from a single prompt to an iterative agentic workflow, and choose the right level of autonomy;
build the Reflection pattern with a Generator, a Critic and external feedback;
give a model tools through function calling, run generated code safely, and meet the Model Context Protocol;
measure an agent with evaluation matrices, spreadsheet error analysis, latency and unit-cost profiling;
design dynamic plans, use Code as Action, and compare multi-agent topologies, up to a transactional agent that never leaves the database half-updated.
Every architecture comes with the numbers that justify it and with small, readable Python. The companion labs include 121 unit tests that run offline with mock LLMs, so you can study the mechanisms without an API key.
Every figure follows one color convention: red for inputs, slate gray for LLM inferences, green for deterministic tools.
Who it is for
Graduate students, engineers and technical leads who build systems around large language models and want to understand how those systems are evaluated, secured and made reliable.
About the Author
Pr. CHEMLAL Yman is a Professor of AI Systems Engineering at the Faculty of Sciences Ben M’Sik, Hassan II University of Casablanca.
Details
Architectures, Design Patterns, and Production-Grade Systems · First Edition