دانلود کتاب ساخت برنامه‌های هوش مصنوعی مولد با Amazon Bedrock

Generative AI crossed a threshold in 2022 when Large Language Models (LLMs) moved from research demos to everyday tools, and the conversation shifted from “Can we build this?” to “How do we ship this safely and reliably?” That shift matters because the hard part is no longer the existence of capable models—it is everything around them such as choosing the right model for a task, controlling cost and latency, keeping customer data protected, preventing prompt injection when models are connected to tools, and building evaluation loops that keep quality stable as requirements evolve. In other words, generative AI is increasingly an engineering discipline, not a novelty feature. Amazon Bedrock fits into this moment as an infrastructure layer for foundation models. Instead of managing model hosting, scaling, and provider-specific integration details yourself, Bedrock gives you access to multiple high-performing foundation models through a managed service and unified interfaces, with an emphasis on security, privacy, and responsible AI capabilities suitable for enterprise use. The goal of this book is to help you use that infrastructure well—to move from “prompting in a playground” to building production-grade systems that behave predictably under real load, integrate with the existing workflows, and remain maintainable over time. A recurring theme, you will see throughout the chapters, is that generative systems fail differently from traditional software. They are probabilistic, and their performance depends on prompts and context length, as well as their costs scale with tokens in ways that can surprise teams after launch. Many of the most common failures are not “model failures” but product and systems failures: unclear success criteria, premature complexity, and shipping, without a credible evaluation strategy. This book, therefore treats prompts, retrieval, tool use, monitoring, and governance as first-class engineering concerns—because they are.