Ashford Partners
LLM Systems Engineering

LLM Systems Engineering

Training and Building Large Language Models — Engineering AI Models Through Fine-Tuning, Continued Pretraining, and From-Scratch Development

AuthorRob McKinseyPages359 pagesPublisherAshford PartnersReleased2026ISBN978-3952663240

About this book

Large language models have become the defining technology of the current decade, powering tools that write and reason at scale. However, there is much more to LLMs than just the chat interface of ChatGPT or Claude: a deeper opportunity to design and adapt models to your own specific needs.

Research scientist and author Rob McKinsey presents a detailed resource on the topic, spanning 42 modules organized into eight major chapters. The discussion begins with the inner workings of an LLM and progresses to data engineering and training production models with Python and modern engineering tools.

Whether you are learning the fundamentals of LLM systems or training and adapting models in professional environments, this book provides a structured and practical guide to understanding and designing language models that perform reliably across real-world workflows.

What you'll learn

  • How tokens, embeddings, and the transformer architecture turn text into model behavior
  • How pretraining objectives and inference loops actually work
  • How to plan data and compute budgets for an LLM project
  • How to fine-tune, continue pretraining, and train from scratch with Python and Unsloth

Table of contents

  1. 1Chapter I — Introduction to Large Language Models
  2. 2Chapter II — Strategic and Technical Planning
  3. 3Chapter III — Data Engineering
  4. 4Chapter IV — Training Systems and Model Development
  5. 5Chapter V — Fine-Tuning
  6. 6Chapter VI — Continued Pretraining
  7. 7Chapter VII — Training from Scratch
  8. 8Chapter VIII — Evaluation, Monitoring, and Maintenance

Who this book is for

This book is written for developers, AI architects, and enthusiasts who want a deeper, foundational understanding of how large language models are composed and trained. It assumes comfort with general programming and basic ideas from linear algebra and probability. The material is presented at an intermediate to advanced level.