Building Agent-Powered Applications

Build Production-Ready Generative AI Systems

Building Agent-Powered Applications is a practical, engineering-focused guide to turning LLM ideas into dependable products. Rather than treating prompting, retrieval, fine-tuning, and agents as separate buzzwords, it shows how these pieces fit together when you are building systems that need to work in the real world.

Written by Vasyl Zvarydchuk, the book starts with the foundations: AI, NLP, embeddings, transformers, and the behavior of large language models. From there, it moves into prompt engineering and core language tasks such as summarization, classification, extraction, and reasoning. The result is a clear path from concepts to implementation, with a strong emphasis on production trade-offs.

What the book covers

• How LLMs, embeddings, and transformers fit into modern AI applications
• Prompting techniques that help produce more reliable outputs
• RAG pipelines for grounding models in useful external information
• When to choose prompting, retrieval, or fine-tuning
• Agent design with tools, memory, planning, and orchestration
• Evaluation methods for quality, reliability, and responsible AI

Why it stands out

This is a book for readers who want more than theory. Its focus on architecture, evaluation, and system behavior makes it especially relevant for developers and technical teams building applications with LLMs and AI agents. The material is organized to support practical decision-making, not just experimentation.

Who will find it useful

It will be a strong fit for AI engineers, software engineers, data scientists, applied AI practitioners, technical leads, and product managers working close to engineering teams. If you are moving from traditional software or classical ML into generative AI, the book offers a structured way to make that transition.

A sensible next step for serious AI builders

If you want a grounded, production-minded ebook on modern LLM applications, this title brings together the key ideas shaping agentic AI today — and explains how to use them with intention.