AI Systems Performance Engineering

□ I. THÔNG TIN SẢN PHẨM
□ Mã sản phẩm : STT1707
□ Nhà xuất bản : O'Reilly Media
□ Tác giả : Chris Fregly
□ Ngôn ngữ : Tiếng Anh
□ ISBN : 9798341627789
□ Số trang : 1058 trang
□ Hình thức : Bìa Mềm, RUỘT IN ĐEN TRẮNG HOẶC MẦU LASER, BÌA IN MẪU LASER GIẤY C300 CÓ CÁN
□ Loại : Sách gia công đóng gáy keo chắc chắn chất lượng cao
□ Giấy in : Giấy ngoại định lượng 70msg, viết vẽ và highlight thoải mái.
□ Chất lượng : Bản in rõ nét, giá rất tốt cho mọi người.

□ II. MÔ TẢ SẢN PHẨM
□ 1.Mô tả sản phẩm đầy đủ
AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch is the definitive resource for engineers, researchers, and developers seeking to maximize the efficiency of their AI infrastructure. In an era dominated by ever-growing generative models and massive datasets, the performance of AI systems has become a critical bottleneck. This book addresses these challenges by providing a hands-on, comprehensive set of actionable optimization strategies that cover every layer of the AI stack. Author Chris Fregly, a seasoned performance expert, guides readers through the complex process of co-optimizing hardware, software, and algorithms to build resilient, scalable, and cost-effective AI solutions. The book begins by establishing a foundational understanding of AI system architecture and the key metrics that define performance in both training and inference. From there, it delves into deep technical territory, offering step-by-step methodologies for fine-tuning GPU CUDA kernels and PyTorch-based algorithms. Readers will learn how to leverage the latest hardware features and compiler-driven accelerations to extract every ounce of throughput from their chips. A significant portion of the text is dedicated to scaling, covering multinode training environments and the intricacies of high-throughput inference servers. Fregly explains how to scale GPU clusters effectively, manage distributed model training jobs, and eliminate the common performance bottlenecks that plague large-scale AI pipelines. Beyond technical tuning, the book emphasizes the economic and operational aspects of AI engineering, showing how optimized systems lead to substantial cost savings and reduced latency in real-world settings. One of the most valuable features is a massive, 175-item checklist of proven, ready-to-use optimizations that can be applied immediately to production workloads. Whether you are dealing with kernel-level tuning or high-level system design, this book provides the tools and insights needed to transform complex AI systems into streamlined, high-impact technologies that excel in the most demanding enterprise environments.

□ 2. Tác giả
Chris Fregly is a performance-focused engineering and product leader who has driven major innovations at companies including Netflix, Databricks, and Amazon Web Services (AWS). As an industry expert in AI/ML infrastructure, he has led teams in building high-scale products and significantly reducing operational costs for massive generative AI and analytics workloads. Chris is the co-author of the popular O’Reilly books 'Data Science on AWS' and 'Generative AI on AWS,' and he created the O’Reilly course 'High-Performance AI in Production with NVIDIA GPUs.' His technical work spans kernel-level tuning, compiler-driven acceleration, distributed training, and high-throughput inference. He is also the organizer of the global AI Performance Engineering meetup, which serves over 100,000 members worldwide.