Building AI Systems with Python: Practical Machine Learning and Agentic Workflows with Python and Pytorch
暫譯: 使用 Python 建構 AI 系統：實用機器學習與 Python 及 Pytorch 的代理工作流程

This book is a practical, end-to-end guide for engineers and practitioners who want to move beyond prototypes and confidently deploy machine learning and large language model solutions in real-world environments.

This book guides through the entire modern machine learning lifecycle. You'll start with foundations using NumPy, Pandas, and PyArrow for data pipelines, then build solid baselines with scikit-learn. From there, you advance into deep learning with PyTorch, transformers, and LLM adaptation techniques such as LoRA and QLoRA. You'll explore diffusion and multimodal models, and learn to build retrieval-augmented generation systems with FAISS and pgvector. Practical chapters cover agents, tool use, evaluation frameworks, observability, and responsible AI practices including privacy, safety, and governance. Finally, you'll master deployment techniques using FastAPI, Ray Serve, TorchServe, and cutting-edge LLM serving engines like vLLM and TGI. Each concept is paired with clear code examples, testing patterns, and operational checklists. Instead of one-off tricks, you'll adopt repeatable workflows: schema-first tooling, reproducible training pipelines, evaluation with golden datasets, and secure production rollouts with monitoring and compliance checkpoints.

In the end, this book helps you build systems that are robust, auditable, and optimized--whether you're deploying your first model or managing complex enterprise workloads. For engineers who want to ship AI confidently and responsibly, this is your practical playbook for the GenAI era.

What you will learn:

Implement modern AI models including transformers, diffusion, multimodal, recommenders, and RL using practical PyTorch examples.

Fine tune and serve LLMs with LoRA/QLoRA, quantization, RAG, tool calling, structured prompts, and robust evaluation techniques.

Design agentic AI systems with memory, planning, safe tool execution, multi agent patterns, and autonomy evaluation frameworks.

Deploy and run production grade AI with MLOps/LLMOps covering serving, performance tuning, monitoring, cost control, compliance, and edge deployments.

Who this book is for:

This book is designed for practicing machine learning and AI engineers, software engineers moving into applied AI, data scientists building production systems, MLOps/LLMOps practitioners, and technical builders who want to go beyond demos and deploy real-world GenAI, LLM, and PyTorch-based systems at scale.