How to use this

The skill tree maps out the conceptual dependencies on the path to understanding large language models. Each node represents a topic; nodes unlock in order, so you can track your own progress through the curriculum.

• Click any node to see a summary and the key sub-topics
• Mark Complete to unlock dependent nodes
• Green glow = available to learn now

The five tiers

Tier 1 — Foundations are always unlocked: linear algebra, calculus, probability & stats, and Python/NumPy. Everything else traces back to these.

Tier 2 — Core ML covers the classical toolkit: supervised and unsupervised learning, optimisation, and model evaluation. If you’ve taken an introductory ML course, most of this will be review.

Tier 3 — Deep Learning starts with neural networks, then branches into CNNs, RNNs/Seq2Seq, and practical frameworks (PyTorch/JAX). RNNs are worth spending real time on — understanding why they struggle with long sequences directly motivates the attention mechanism.

Tier 4 — Transformers is the conceptual heart. The attention mechanism comes first, then the full transformer architecture.

Tier 5 — LLMs builds on transformers to cover tokenisation, pretraining & scaling laws, fine-tuning & RLHF, and inference & prompting.