# The Craft of Post-Training: A Practical Guide for AI Engineers and Developers

> Post-training is the process of adapting a pre-trained foundation model to specific tasks, domains, or behavioural requirements through techniques such as supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), and direct preference optimisation (DPO). Unlike prompting or retrieval-augmented generation (RAG), post-training permanently modifies the model's weights to embed new capabilities, domain knowledge, or alignment constraints. This book is a practical guide to turning foundation models into production-ready systems. By Chris von Csefalvay, published by No Starch Press in 2026 (ISBN 9781718505209, 416 pages) and now shipping on Amazon.

## Book information
- Title: The Craft of Post-Training: A Practical Guide for AI Engineers and Developers
- Author: Chris von Csefalvay
- Publisher: No Starch Press
- ISBN: 9781718505209
- Pages: 416
- Edition: 1st Edition
- Publication: 2026, No Starch Press
- Status: Released and shipping on Amazon
- Homepage: https://posttraining.guide
- Buy: https://www.amazon.com/dp/1718505205
- No Starch Press: https://nostarch.com/craft-of-post-training
- Penguin Random House: https://www.penguinrandomhouse.com/books/851009/the-craft-of-post-training-by-chris-von-csefalvay/
- Google Books: https://books.google.com/books/about/The_Craft_of_Post_Training.html?id=dUXbEQAAQBAJ
- Chapter 3 sample: https://nostarch.com/download/samples/craft-of-post-training-chapter-3.pdf
- Companion notebooks: https://github.com/chrisvoncsefalvay/craft-of-post-training/tree/main/notebooks
- Companion skill: https://posttraining.guide/post-training-guide-skill.md
- Author Website: https://chrisvoncsefalvay.com

## What you'll learn
1. Fine-tune models on curated datasets using supervised fine-tuning (SFT), Low-Rank Adaptation (LoRA), and Quantised LoRA (QLoRA) without destroying the base model's general capabilities.
2. Apply reinforcement learning from verifiable rewards (RLVR) and modern preference optimisation methods including Direct Preference Optimisation (DPO), Odds Ratio Preference Optimisation (ORPO), and beyond, to shape model behaviour.
3. Evaluate models rigorously: design benchmarks, detect regression, and measure quality claims that survive scrutiny.
4. Adapt models to specialised domains — from clinical language to legal text — turning general capability into a defensible competitive advantage.
5. Train agentic models that take sequences of actions reliably, not just models that talk about taking actions.
6. Quantise and compress fine-tuned models for deployment without sacrificing the gains you trained for.

## Why this book
The literature on post-training is focused either on small educational use cases that do not consider enterprise realities, or presuppose the workflow of foundation labs. There's nothing for the crucial middle: enterprise practitioners with real compute budgets who need to customise, align and deploy AI at scale. This book fills that gap. It treats post-training decisions as trade-offs rather than best practices, helping practitioners match techniques to constraints. It combines technical depth with strategic context and includes companion Jupyter notebooks covering practical implementation.

## Table of contents

### Part I: The Foundation
- Chapter 1: Post-Training Essentials: What It Is and Why It Matters
- Chapter 2: Prerequisites for Success: Before You Fine-Tune

### Part II: The Tools
- Chapter 3: Supervised Fine-Tuning: The Foundation Technique
- Chapter 4: Reinforcement Learning: Better Each Time
- Chapter 5: Preference Optimization: Modern Alternatives to PPO
- Chapter 6: Evaluation Strategies: Measuring Model Quality

### Part III: The Craft
- Chapter 7: Efficiency Techniques: Quantization and Compression
- Chapter 8: Domain Adaptation: Make It Yours
- Chapter 9: Agentic Models: Deeds, Not Words
- Chapter 10: Reasoning Capabilities: Training for Complex Thought

### Part IV: The Frontier
- Chapter 11: Synthetic Training: Self-Play and Generated Data
- Chapter 12: Multimodal Systems: Post-Training Beyond Text
- Chapter 13: Future Directions: What Comes Next

## Topics covered
- Supervised fine-tuning (SFT) with Low-Rank Adaptation (LoRA) and Quantised LoRA (QLoRA)
- Reinforcement learning from verifiable rewards (RLVR)
- Group Relative Policy Optimisation (GRPO)
- Direct Preference Optimisation (DPO), Kahneman-Tversky Optimisation (KTO), Odds Ratio Preference Optimisation (ORPO)
- Model evaluation, benchmark design, and regression detection
- Domain adaptation and specialisation (clinical, legal, enterprise)
- Agentic model training for sequential action-taking
- Quantisation, compression, and deployment optimisation
- Multimodal and vision-language alignment
- Synthetic data generation and RLAIF
- Reasoning capabilities and complex thought training

## Frequently asked questions

### What is post-training?
Post-training is the process of adapting a pre-trained foundation model to specific tasks, domains, or behavioural requirements. Unlike prompting or RAG, post-training permanently modifies the model's weights through techniques such as supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), and direct preference optimisation (DPO).

### What is the difference between fine-tuning and post-training?
Fine-tuning (specifically supervised fine-tuning or SFT) is one technique within the broader post-training toolkit. Post-training also includes reinforcement learning methods (RLHF, RLVR, PPO), preference optimisation (DPO, ORPO, KTO), domain adaptation, agentic training, and evaluation — the full lifecycle of turning a foundation model into a production-ready system.

### Who is this book for?
Enterprise AI engineers and ML practitioners who need to customise, align, and deploy language models at scale. The book assumes familiarity with Python, PyTorch, and the fundamentals of machine learning, but does not require prior experience with post-training specifically.

### What post-training techniques does the book cover?
The book covers supervised fine-tuning (SFT) with LoRA and QLoRA, reinforcement learning from verifiable rewards (RLVR) with GRPO, preference optimisation (DPO, ORPO, KTO), model evaluation and benchmark design, domain adaptation, agentic model training, quantisation, multimodal alignment, and synthetic data generation.

### Do I need access to GPUs to follow along?
Many examples use parameter-efficient methods like LoRA that can run on a single consumer GPU. The book targets enterprise practitioners with real but manageable compute budgets, not foundation model labs with thousands of GPUs.

## Author
Chris von Csefalvay is a Principal at HCLTech's AI Practice, where he leads post-training research and clinical intelligence. He has held senior data science leadership roles across major enterprises, published extensively on distributed computing for ML, and designed language models for applications ranging from pharmacovigilance to social dynamics. He is the author of Computational Modeling of Infectious Disease (Elsevier, 2023). He holds degrees from the University of Oxford and Cardiff University and is a Fellow of the Royal Society for Public Health and Senior Member of IEEE.

- ORCID: https://orcid.org/0000-0003-3131-0864
- Google Scholar: https://scholar.google.com/citations?user=X_2G-VsAAAAJ
- Wikidata: https://www.wikidata.org/wiki/Q107095298
- LinkedIn: https://www.linkedin.com/in/chrisvoncsefalvay/
- GitHub: https://github.com/chrisvoncsefalvay