Free · interactive · no account

Learn machine learning from the math up to AI engineering, by touching it.

108 visual explainers you can drag and tune, arranged as one path: from derivatives to transformers to serving LLMs in production.
FIG. 01 · SIX TRACKS, ONE PATH01 Foundations02 Core ML03 Deep Learning04 LLMs05 AI Engineering06 System Design
Where are you starting from?

Pick the door that sounds like you

Each door opens the first explainer of a learning path. You can switch any time, and every door leads into the same journey below.

Want the math first? Start with Foundations, or dip into it from any track.

The journey

Six tracks, one ladder

Each track builds on the one before it. Start wherever fits; the concept map shows every prerequisite.

1

Foundations

The math under machine learning

You'll be able to: Read the derivatives, vectors, and probabilities that every model is written in.

DerivativesThe Chain RuleEigenvalues & EigenvectorsBayes' Theorem
3 paths · 20 explainers
Open track →
2

Core ML

How models learn from data

You'll be able to: Train and evaluate a model, and tell overfitting from a real result.

Gradient DescentDecision Trees & SplittingCross-ValidationBias-Variance Tradeoff
4 paths · 27 explainers
Open track →
3

Deep Learning

Neural networks, end to end

You'll be able to: Explain backpropagation, and why deep networks need careful initialization and normalization.

BackpropagationOptimizersConvolutional Neural NetworksSkip Connections & ResNets
4 paths · 25 explainers
Open track →
4

LLMs & Generative AI

From words as numbers to GPT

You'll be able to: Read a transformer diagram and explain how GPT, diffusion, and RAG work inside.

EmbeddingsHow Self-Attention WorksThe Transformer ArchitectureDiffusion Models
4 paths · 22 explainers
Open track →
5

AI Engineering

Training, serving, and running models in production

You'll be able to: Reason about the speed, memory, and cost of training and serving LLMs.

KV CacheQuantizationFlashAttentionVector Search & HNSW
3 paths · 16 explainers
Open track →
6

AI System Design

Putting the parts together

You'll be able to: Design whole AI systems under real limits on latency, cost, and quality.

Serve an LLMBuild a RAG SystemDesign a Recommendation Feed
3 interactive case studies
Next case studies: building a RAG system and designing a recommendation feed.
Open track →

Browse all 18 learning paths →

Bites

Latest bites

Comparison
bf16 vs fp16: range or precision in the same 16 bits
2 min read
Comparison
Continuous vs static batching: why LLM servers refill slots every step
2 min read
Comparison
MoE vs dense: add experts, not compute per token
2 min read
Comparison
Adam vs SGD: two ways down the same hill
2 min read
Comparison
Bagging vs boosting: many trees side by side or one after another
2 min read
Plain words
Batch normalization in plain words
2 min read
Free for everyoneNo account neededEvery concept is interactive108 explainers · 18 paths