Skip to main content

CONCEPT-FIRST LEARNING LIBRARY

Build the mental model.
Then build the systems.

LLM Basics turns a dense certification map into a calm, practical curriculum: theory, tradeoffs, mini-labs, and the resources worth your time.

Begin with LLM fundamentals ↗See the study path
10
connected chapters
7–10h
per foundation block
1 goal
understand before testing

THE APPROACH

No acronym collecting.

You will learn to choose an approach, predict its tradeoffs, and diagnose what broke. That is the difference between recognizing a term and being able to use it.

THE LEARNING MAP

Ten chapters. One connected system.

Start where every practical decision starts: with how a model represents and generates language. Then move outward—data, adaptation, systems, operations, and safety.

  1. 01LLM fundamentalsTokens, embeddings, attention, architectures, and decoding.Explore ↗
  2. 02Prompting + RAGMake model behavior useful, grounded, and testable.Explore ↗
  3. 03Data preparationBuild clean, attributable data pipelines and splits.Explore ↗
  4. 04OptimizationQuantization, distillation, KV cache, and real tradeoffs.Explore ↗
  5. 05Fine-tuningSFT, LoRA, preference learning, and alignment.Explore ↗
  6. 06EvaluationMeasure quality, retrieval, reliability, and regressions.Explore ↗
  7. 07GPU systemsMemory, parallelism, precision, and profiling.Explore ↗
  8. 08DeploymentServing, batching, orchestration, and rollout.Explore ↗
  9. 09ReliabilityObservability, incident response, and versioning.Explore ↗
  10. 10SafetyGuardrails, risk, privacy, and responsible AI.Explore ↗

START HERE

Understand attention before you optimize it.

Chapter 1 gives you the vocabulary and intuition behind tokens, embeddings, transformer blocks, self-attention, model families, decoding, and KV cache.

Open Chapter 1 ↗