Books
The lessons follow these books. Each lesson ends with the chapters to read for more depth; this page lists them the other way round, book by book.
Deep Learning
Goodfellow, Bengio, Courville, 2016 · free online ↗
Vectors & dot products · Matrices & matrix multiplication · Probability & softmax · Logs, cross-entropy & perplexity · Derivatives & the chain rule · Mean, variance & normalisation · Data & features · Linear regression · Gradient descent · Overfitting · Neural nets & backprop · Optimisers · n-grams to neural LMs · Why transformers · FFN, GLU & normalisation
Dive into Deep Learning
Zhang, Lipton, Li, Smola, 2023 · free online ↗
Vectors & dot products · Matrices & matrix multiplication · Probability & softmax · Logs, cross-entropy & perplexity · Derivatives & the chain rule · Linear regression · Gradient descent · Classification · Overfitting · Neural nets & backprop · Optimisers · n-grams to neural LMs · Embeddings · Attention · Transformer block · Why transformers · Multi-head attention & masks · Positional encoding · Encoder, decoder, or both
Build a Large Language Model (From Scratch)
Sebastian Raschka, 2024
Mean, variance & normalisation · Neural nets & backprop · Optimisers · Tokenisation · Attention · Transformer block · Sampling · Multi-head attention & masks · Positional encoding · FFN, GLU & normalisation · Pretraining & scaling laws · Fine-tuning
AI Engineering
Chip Huyen, 2025
Logs, cross-entropy & perplexity · Sampling · KV cache, MQA & GQA · Mixture of experts · LLM families & Chinchilla · Multimodal LLMs · Pretraining & scaling laws · Fine-tuning · RLHF & DPO · KV cache & quantisation · Benchmarks & metrics · Precision, recall & F1 · BLEU, ROUGE & BERTScore · Ranking: MRR & nDCG · Contamination · LLM-as-judge · Human eval & arenas · Hallucination & calibration · Evaluating RAG & agents · Latency, throughput & cost · Prompting & context · RAG · Structured output · LangChain building blocks · Tools with @tool · The agent loop · Middleware & guardrails · Memory: threads & checkpoints · MCP tools · Agent security · Workflow patterns
Hands-On Large Language Models
Jay Alammar, Maarten Grootendorst, 2024
Tokenisation · Embeddings · Attention · Transformer block · Sampling · Positional encoding · Encoder, decoder, or both · KV cache, MQA & GQA · LLM families & Chinchilla · Multimodal LLMs · RLHF & DPO · KV cache & quantisation · Prompting & context · RAG
Reinforcement Learning from Human Feedback
Nathan Lambert, 2026 · free online ↗
Speech and Language Processing (3rd ed. draft)
Jurafsky, Martin, 2025 · free online ↗
Tokenisation · n-grams to neural LMs · Embeddings · BLEU, ROUGE & BERTScore
Pattern Recognition and Machine Learning
Christopher Bishop, 2006
Hands-On Machine Learning with Scikit-Learn, Keras and TensorFlow (3rd ed.)
Aurélien Géron, 2022
Data & features · Gradient descent · Classification · Precision, recall & F1
Designing Machine Learning Systems
Chip Huyen, 2022
Introduction to Information Retrieval
Manning, Raghavan, Schütze, 2008 · free online ↗
Benchmarks & metrics · Precision, recall & F1 · Ranking: MRR & nDCG · Evaluating RAG & agents · RAG
Artificial Intelligence: A Modern Approach (4th ed.)
Russell, Norvig, 2020
Tools with @tool · The agent loop · Memory: threads & checkpoints · Handoffs between agents · Choosing a multi-agent pattern
An Introduction to MultiAgent Systems (2nd ed.)
Michael Wooldridge, 2009
Mathematics for Machine Learning
Deisenroth, Faisal, Ong, 2020 · free online ↗
Vectors & dot products · Matrices & matrix multiplication · Probability & softmax · Derivatives & the chain rule · Mean, variance & normalisation
Learning LangChain
Mayo Oshin, Nuno Campos, 2025
LangChain building blocks · State, nodes & edges · Routing & loops · Checkpoints & human approval
LangChain and LangGraph documentation
LangChain, 2026 · free online ↗
LangChain building blocks · Tools with @tool · The agent loop · Middleware & guardrails · Memory: threads & checkpoints · MCP tools · Agent security · State, nodes & edges · Routing & loops · Workflow patterns · Parallel work with Send · Checkpoints & human approval · Supervisor & subagents · Handoffs between agents · Choosing a multi-agent pattern
TinyML and Efficient Deep Learning Computing, Lecture 12: Transformer and LLM (slides)
Song Han, MIT 6.5940, 2024 · free online ↗
Why transformers · Multi-head attention & masks · Positional encoding · Encoder, decoder, or both · KV cache, MQA & GQA · FFN, GLU & normalisation · Mixture of experts · LLM families & Chinchilla · Multimodal LLMs