SK / ENGINEERING BLOG · 2026 EDITION
From Machine Learning
to Transformers.
A step-by-step illustrated reference starting at absolute zero, moving through classical algorithms and neural networks, and ending with BERT, generative models, efficient attention and production ML systems.
Start reading →Download all 10 chapters (ZIP) ↓ (online)Contents · 10 chapters
How Machines Learn from Data
Before asking how a neural network thinks, we need a clearer question: what does it mean for any program to learn? This chapter starts with familiar…
CHAPTER 02 · 10 SECTIONS · ~12 MINClassical Machine Learning, Explained
Before deep learning, many remarkably capable models were built from lines, distances, probabilities, and decision rules. These methods remain pract…
CHAPTER 03 · 10 SECTIONS · ~12 MINNeural Networks from the First Neuron
Neural networks are not mysterious digital brains. They are compositions of simple numerical operations whose parameters are adjusted using data. Th…
CHAPTER 04 · 10 SECTIONS · ~12 MINDeep Learning Architectures and Representations
The basic training machinery stays the same, but the arrangement of layers matters. Images, text, sound, and time series have different structures. …
CHAPTER 05 · 10 SECTIONS · ~13 MINNatural Language Processing from Raw Text
A computer receives text as characters and bytes, not as meaning. This chapter follows a sentence through the classical NLP pipeline: cleaning, spli…
CHAPTER 06 · 10 SECTIONS · ~12 MINFrom Word Counts to Language Understanding
Earlier pipelines counted words and matched rules. This chapter introduces vector representations that are *learned* from text, then shows how recur…
CHAPTER 07 · 10 SECTIONS · ~12 MINTransformers: Attention All the Way Down
Transformers are easier to understand if you remember a single idea: every token should be able to gather relevant information from other tokens wit…
CHAPTER 08 · 10 SECTIONS · ~11 MINBERT, GPT, T5, and the Major Transformer Families
A Transformer is an architecture, not a single model or task. This chapter explains how changing the attention direction, training objective, and ou…
CHAPTER 09 · 10 SECTIONS · ~12 MINAdvanced Transformers: Efficient Attention and Modern Architectures
Large Transformer models introduce two costs that a beginner diagram can hide: computation across many token pairs and memory to hold parameters, ac…
CHAPTER 10 · 10 SECTIONS · ~12 MINBuilding and Evaluating Real ML and Language Systems
Knowing architectures is not enough to ship a reliable system. The final chapter joins datasets, training, retrieval, generation, deployment, and mo…
How to use this blog
Read chapters in order: each section introduces its own terminology, provides concrete examples, and explains important trade-offs. No exercises, quizzes, or embedded activities. Every chapter is also available as a complete, print-friendly PDF with selectable text and original diagrams.
Source note. Built with guidance from your seven uploaded PDF references and two companion cheatsheets, reviewed against primary papers and maintained documentation. Notes are synthesized as original explanations. Uploaded PDF pages and diagrams are not redistributed; all included vector illustrations are newly authored.
Selected primary references
- scikit-learn: Common pitfalls
- scikit-learn: Model evaluation
- scikit-learn: Supervised learning
- scikit-learn: Unsupervised learning
- PyTorch: Autograd and computational graphs
- Deep Learning book: Optimization
- Deep Residual Learning for Image Recognition
- Sequence to Sequence Learning with Neural Networks
- spaCy: Linguistic features
- Hugging Face: Tokenizers pipeline
- Word2Vec: Efficient Estimation of Word Representations
- GloVe: Global Vectors for Word Representation
- Neural Machine Translation by Jointly Learning to Align and Translate
- Vaswani et al.: Attention Is All You Need
- Devlin et al.: BERT
- Raffel et al.: T5
- Hugging Face: Transformers documentation
- Su et al.: RoFormer / RoPE
- Ainslie et al.: GQA
- Dao et al.: FlashAttention
- Kwon et al.: PagedAttention
- DeepSeek-V2: MLA
- Hu et al.: LoRA
- Lewis et al.: RAG
- Rafailov et al.: DPO