Hands-On Large Language Models
I keep this at #36 because a visual and practical technical foundation for understanding and using modern language models. I have not finished learning what it asks of me.
The long version
Alammar and Grootendorst build an illustrated, code-oriented path from transformer internals to embeddings, tokenization, text generation, semantic search, clustering, fine-tuning, retrieval, and multimodal systems. Visual explanations make abstract mechanisms tangible, while notebooks connect them to current open-source model workflows. I keep coming back because the question becomes more honest when it reaches my own life.
Why it is here
It is the best technical bridge on the shelf between conceptual AI literacy and hands-on experimentation. It helps you see what model representations and generation loops are doing before you make product or infrastructure decisions around them. I want to carry one useful thing from it into the next difficult moment.
How to read it
Run each notebook in a controlled environment and annotate tensor shapes, inputs, outputs, and evaluation assumptions. After every technique, write one failure case and one production concern. Expect libraries to change; preserve concepts separately from APIs. I will mark the places that make me defensive or relieved.
What it taught me
- 01
Transformers build contextual representations through attention and layered computation.
- 02
Embeddings enable semantic comparison, search, clustering, and retrieval workflows.
- 03
Model choice, prompting, fine-tuning, and RAG correspond to different engineering needs.
Before and after
What makes it easier
Quick Start Guide to Large Language Models
Sinan Ozdemir
The Illustrated Transformer
Jay Alammar
Where it leads
AI Engineering: Building Applications with Foundation Models
Chip Huyen
Natural Language Processing with Transformers
Lewis Tunstall; Leandro von Werra; Thomas Wolf
More about the author
Jay Alammar is known for visual explanations of machine learning, especially transformers and language models. Maarten Grootendorst is a data scientist and author focused on interpretable NLP, embeddings, clustering, and topic modeling.
More by the same hand
The Illustrated Transformer
Jay Alammar
Hands-On Large Language Models
Jay Alammar; Maarten Grootendorst