{ "item_title" : "LLM Inference Engineering in Production", "item_author" : [" Bevis Thibault "], "item_description" : "LLM Inference Engineering in Production: Deploy Open Models, Optimize GPU Costs, and Scale Reliable AI Services with vLLM and KubernetesRunning an open-source LLM is one thing. Keeping it fast, reliable, and cost-efficient under production traffic is another.Your inference server works in development, but what happens when concurrent requests increase, GPU memory fills up, and response times deteriorate?For AI engineers, MLOps professionals, DevOps engineers, and Kubernetes platform developers, production inference introduces difficult decisions about latency, throughput, GPU utilization, security, and infrastructure spending.Without effective optimization and monitoring, even powerful GPUs can deliver disappointing performance and unnecessary operating costs.LLM Inference Engineering in Production provides a practical, implementation-focused approach to deploying, benchmarking, optimizing, and scaling open-weight language models using vLLM and Kubernetes.Following a continuous engineering project, this book takes you from your first GPU-accelerated inference endpoint to a monitored, secure, and scalable serving architecture.Through hands-on examples, Python implementations, Kubernetes configurations, and technical diagrams, you'll learn how to: Deploy open models confidently: Configure vLLM, build OpenAI-compatible inference endpoints, and deploy GPU-accelerated services on Kubernetes.Improve inference performance: Measure TTFT, throughput, and latency while optimizing continuous batching, PagedAttention, chunked prefill, and prefix caching.Reduce GPU memory usage and costs: Apply quantization, KV-cache optimization, speculative decoding, and capacity-planning techniques.Build reliable production services: Implement authenticated gateways, Prometheus monitoring, Grafana dashboards, health checks, and KEDA-based autoscaling.Scale across multiple GPUs: Understand tensor parallelism, pipeline parallelism, intelligent request routing, distributed inference, and KV-cache offloading.Designed for developers and infrastructure professionals with foundational Python, Docker, or cloud experience, the book connects essential inference concepts to practical production engineering decisions.Effective inference engineering depends on measurable results, reproducible configurations, and reliable operations-not theoretical performance claims.Across 12 structured chapters, the book develops a progressive production inference platform, connecting deployment, GPU optimization, benchmarking, Kubernetes orchestration, observability, security, and cost analysis.You'll work with practical configurations, performance-testing procedures, architecture diagrams, troubleshooting techniques, and production-readiness checklists.Each major optimization follows a disciplined approach: establish a baseline, implement a change, measure its effects, and evaluate whether the improvement justifies its cost and complexity.By the end, you'll understand how to diagnose inference bottlenecks, evaluate GPU capacity, improve serving efficiency, and make informed infrastructure decisions.Move beyond simply running language models. Build the engineering skills to deploy, optimize, and operate production inference systems with confidence.Get your copy of LLM Inference Engineering in Production and start building faster, more reliable, and cost-conscious AI infrastructure.", "item_img_path" : "https://covers3.booksamillion.com/covers/bam/9/79/816/725/9798167250062_b.jpg", "price_data" : { "retail_price" : "27.00", "online_price" : "27.00", "our_price" : "27.00", "club_price" : "27.00", "savings_pct" : "0", "savings_amt" : "0.00", "club_savings_pct" : "0", "club_savings_amt" : "0.00", "discount_pct" : "10", "store_price" : "" } }

Overview

Customers Also Bought

Details