Back to AI/ML Overview

Large Language Models

How ChatGPT, Claude, and modern AI systems work

Beginner to Intermediate
10 of 10 lessons available
3-4 hours total
1

What LLMs Are

Understand the foundations of large language models

Start Here
Beginner10-12 min

What is an LLM?

From simple language models to GPT-4 — understand what makes a language model 'large' and why it matters.

Language ModelingScaleCapabilitiesLimitations
Start Learning
Beginner10-12 min

Tokenization: Text to Numbers

See how text gets split into tokens and converted to numbers — the first step in any LLM pipeline.

BPESubword TokensVocabularySpecial Tokens
Explore
Intermediate12-15 min

Pre-training: Learning from the Internet

Watch a model learn language patterns from billions of tokens — next-token prediction at massive scale.

Next-Token PredictionCross-EntropyTraining DataCompute
Explore
2

Scale & Generation

How LLMs generate text and why bigger models perform better

Intermediate10-12 min

Scaling Laws: Bigger is Better?

Discover the mathematical relationship between model size, data, compute, and performance.

Chinchilla LawsEmergent AbilitiesCompute-Optimal
Explore
Intermediate12-15 min

Text Generation: Next Token Prediction

Follow the autoregressive loop that turns probability distributions into coherent text.

AutoregressiveProbability DistributionGreedy Decoding
Explore
Intermediate10-12 min

Sampling: Temperature, Top-k, Top-p

Control creativity vs coherence by adjusting how the model picks the next token.

TemperatureTop-kTop-p / NucleusDiversity
Explore
3

Alignment & Control

Making LLMs helpful, harmless, and honest

Intermediate12-15 min

Fine-tuning: Teaching New Tricks

Adapt a pre-trained model to specific tasks with supervised fine-tuning and parameter-efficient methods.

SFTInstruction TuningLoRAQLoRA
Explore
Intermediate12-15 min

RLHF: Learning from Human Feedback

See how human preferences train a reward model that guides the LLM toward helpful, safe responses.

Reward ModelPPODPOAlignment
Explore
4

Practical Usage

Get the most out of LLMs in practice

Beginner10-12 min

Prompt Engineering

Master zero-shot, few-shot, and chain-of-thought prompting to get better results from any LLM.

Zero-shotFew-shotChain-of-ThoughtSystem Prompts
Explore
Intermediate12-15 min

Context Windows & Memory

Understand the attention bottleneck, KV cache, and how RAG extends an LLM's knowledge.

Context WindowKV CacheRoPERAG Preview
Explore

Continue Your Journey

After understanding LLMs, learn how to augment them with external knowledge