Building Agent-Powered Applications: Your guide to generative AI, RAG, fine-tuning, and orchestration for production use

Vasyl Zvarydchuk

Move from experimentation to real-world deployment with LLM and agentic applications powered by prompting, RAG, fine-tuning, and evaluation

Key Features
• Design LLM apps by combining prompting, RAG, fine-tuning, and agents
• Evaluate reliability, quality, and safety across real-world AI workflows
• Build production-ready generative AI systems with practical trade-offs

Large language models can produce impressive demos, but turning them into reliable products takes more than better prompts. You need to understand model behavior, know when to use retrieval or fine-tuning, structure agents correctly, and evaluate systems before deployment.

Building Agent-Powered Applications gives an end-to-end engineering perspective on creating production-ready generative AI solutions. Written by Microsoft Principal AI Engineer Vasyl Zvarydchuk, it helps software engineers, data scientists, and applied AI practitioners move from concept to implementation. You’ll begin with AI, NLP, embeddings, transformers, and LLM behavior, then progress to prompt engineering, summarization, classification, extraction, reasoning, RAG, and fine-tuning.

The book shows how to design agentic workflows with tools, memory, planning, orchestration, and human-in-the-loop controls. You’ll learn to evaluate quality with offline and online testing, task-specific metrics, LLM-as-a-judge methods, and responsible AI checks. Rather than treating prompting, RAG, fine-tuning, and agents as separate topics, this book shows how they work together in practice. By the end, you’ll be able to make better architectural trade-offs, reduce failure modes, and build scalable, trustworthy AI applications.

Who this book is for
This book is for AI Engineers, data scientists, software engineers, applied AI practitioners, technical leads, and engineering-focused product managers who want to build production-ready applications with LLMs and AI agents. It suits readers moving from traditional software developmen