商品詳細

This book provides a practical and rigorous guide to building AI systems whose technical answers can be inspected, checked, and improved. It explains why fluent language generation and retrieval-augmented generation are not enough by themselves and then develops the evidence pipeline needed for trustworthy technical reasoning. Readers learn to treat an AI assistant as a reference-aware system rather than as a model alone. Each chapter turns reliability into concrete questions: What was retrieved? Which claim does each citation support? What assumptions were used? What can be recomputed or tested? What remains uncertain? When should the system abstain? The book is intended for machine learning engineers, data scientists, software engineers, applied researchers, technical managers, and advanced students who need to design or evaluate AI assistants for high-value technical work.

• Provides a practical guide to building AI systems whose technical answers can be inspected, checked, and improved
• Includes evidence-first RAG methodology connecting retrieval design choices to citation fidelity and answer reliability
• Presents claim-level verification with uncertainty-aware response policies for trustworthy technical reasoning