Mastery Libri GmbH Mastery of Generative AI, Volume II: The Practitioner's Depth — From Transformers to Production, A Complete Technical Reference for LLM Engineers

Volume II of Mastery of Generative AI is the engineering practice of building and running large language models in production.

From GPU systems through distributed training, inference optimization, retrieval, deployment, evaluation, and the edge. For engineers shipping LLM systems - everything that happens after the architecture is decided: how to actually run these models at scale, latency, and cost.

Inside Volume II (Chapters 10-21):
- GPU systems and the pivot to neural acceleration
- Memory: KV-cache, FlashAttention, paged attention, gradient checkpointing
- Distributed training: data, tensor, pipeline, and expert parallelism
- Inference optimization: speculative decoding, quantization (INT8, INT4, FP8, AWQ, GPTQ), batching
- Prompt engineering and structured generation
- Retrieval-augmented generation: embeddings, hybrid retrieval, rerankers
- Deployment patterns, containerization, observability
- Fairness, bias, and responsible AI
- Evaluation: benchmarks, LLM-as-judge, detecting silent regressions
- Edge computing and on-device inference
>Aligned with the NVIDIA-Certified Professional: Generative AI LLMs and NVIDIA-Certified Professional: Agentic AI exams. While this book is not an official NVIDIA publication, it covers the public technical foundations these certifications assess.

12 chapters, plus Implementation Guide (Appendix B), Performance Optimization Checklists (Appendix C), Bibliography (Appendix D), Glossary (Appendix E), and Volume II Index (Appendix F).

Volume I - The Conceptual Core - covers the conceptual foundations: transformers, alignment, scaling laws (Chapters 1-9).