Python and the small amount of maths you need before anything else — start here if you have never written code.
2 sections 17 of 17 lessons published
Learn. Build. Ask. Share.
Learn Artificial Intelligence the easy way.
Free lessons on machine learning, deep learning, computer vision, NLP and large language models. Every lesson is written three times over: plain English with no maths, working code you can run, and the mathematics underneath. You pick the depth.
Try: What is machine learning? What is a neural network? What is a large language model? What is RAG? Prompt engineering
Start from zero Browse all topics
No sign-up. No ads. No tracking. Nothing behind a paywall, ever.
One lesson, three depths
The same idea explained for three different readers. Switch between them on any lesson page — your choice is remembered on your device.
-
Beginner
No maths. Plain English.
An everyday analogy first, then the idea in plain words. No symbols, no jargon left undefined.
-
Developer
Code and libraries.
The install line, code that runs as written, the output you should see, and the mistakes people hit.
-
Researcher
Mathematics and papers.
The mathematics with every symbol defined, the cost of the method, and the papers it came from.
Browse by area
All 117 sections14 areas, 117 sections, 1,167 lessons. Start wherever you already are — the areas run in learning order, and so does everything inside them.
-
-
How a machine learns from data, and the classical algorithms that still beat deep learning on most real tables.
- Machine Learning
- Classic Algorithms in Depth
- Linear Models and Regularisation
- Ensembles and Gradient Boosting
- Preprocessing and Feature Selection
- and 6 more
11 sections 105 of 105 lessons published
-
Telling a real result from a lucky one: significance, A/B tests, and working out what actually caused what.
- Hypothesis Testing and Inference
- Resampling, Likelihood and Bayes
- Experiment Design and A/B Testing
- Causal Inference Basics
4 sections 42 of 42 lessons published
-
The judgement calls nobody teaches — what to build, what to compare against, and how to know your numbers are honest.
- Scoping an ML Project
- Baselines and Choosing a Model
- Reading Training Curves
- Data Leakage and Results That Are Too Good
- Reproducibility and Running Experiments
- and 4 more
9 sections 79 of 79 lessons published
-
Neural networks explained in plain words, then built by hand in PyTorch, all the way to training on many GPUs.
- Deep Learning
- PyTorch Tensors
- Autograd in Depth
- Building Models with nn.Module
- Datasets and DataLoaders
- and 6 more
11 sections 113 of 113 lessons published
-
The libraries you will actually type: scikit-learn, HuggingFace, TensorFlow and JAX.
4 sections 40 of 40 lessons published
-
Everything to do with text — cleaning it, embedding it, searching it, and answering questions from it.
- Natural Language Processing
- Text Preprocessing
- Classical NLP That Still Works
- The Embedding Family
- The BERT Family
- and 10 more
15 sections 140 of 140 lessons published
-
Teaching a computer to see: classification, detection, segmentation, faces, video, documents and 3D.
- Computer Vision
- CNN Backbones and Pretraining
- Training Vision Models
- Object Detection in Depth
- Segmentation in Depth
- and 8 more
13 sections 143 of 143 lessons published
-
The other big families of models — audio and speech, forecasting what happens next, and choosing what to show a person.
3 sections 36 of 36 lessons published
-
Models that write, draw and answer questions — and how to build a real product on top of them.
4 sections 44 of 44 lessons published
-
Open the model up: attention, tokenisers, how text is generated, and how these things are trained and shrunk.
- Attention Mechanics
- Inside a Transformer Block
- How Models Know Word Order
- Tokeniser Internals
- How Text Is Generated
- and 6 more
11 sections 115 of 120 lessons published
-
Getting a model off your laptop and keeping it alive: data pipelines, serving, scaling, monitoring, cost and on-device.
- MLOps
- Data Engineering for AI
- Feature and Data Pipelines in Production
- Serving Models in Production
- Batching and Concurrency
- and 11 more
16 sections 157 of 157 lessons published
-
What AI actually looks like in hospitals, banks, shops, factories, farms, schools and laboratories.
- AI in Healthcare
- Medical Imaging AI
- AI in Finance and Fraud
- Retail, Demand and Supply Chain
- Manufacturing and Predictive Maintenance
- and 7 more
12 sections 111 of 111 lessons published
-
Bias, privacy, safety and the rules — the part that decides whether a system should be built at all.
2 sections 20 of 20 lessons published
Follow a path
All pathsIf you do not know what order to learn things in, use one of these. Each path is an ordered list of lessons, start to finish.
Become an ML Developer
From zero programming to training and evaluating real models.
Become an AI Engineer
Build production applications on top of large language models.
Become a Computer Vision Engineer
Work with images and video, from OpenCV to vision transformers.
Just added
Feed10 new lessons every day go live.
-
Scaling and Traffic Management
Autoscaling on the right metric
Autoscaling adds and removes machines automatically, but only helps if it watches the number that actually predicts trouble.
-
Evaluating Text Systems
BLEU, chrF and COMET
BLEU scores a translation by counting matching word chunks against a reference, which is fast but blind to correct paraphrases.
-
Classical NLP That Still Works
Bag of words
Bag of words turns a sentence into a list of word counts, throwing away word order but keeping enough signal to search and classify text cheaply.
-
Messy Real-World Text
Building a spell checker
A spell checker guesses the intended word from a misspelled one by finding the closest real word, the same way you guess a word mumbled in a noisy market.
-
Chunking and Long Documents
Chunking strategies compared
Chunking cuts a long document into smaller pieces so a search system can find and hand over only the part that answers a question, and the cutting method decides whether those pieces still make sense.
-
Medical Imaging AI
DICOM, windowing and image intensity
A medical scan is not a photograph — its pixels store a physical measurement, and you choose which slice of that range to actually look at.
Never written a line of code?
That is the normal starting point here. Do these three things in order and you will have trained your first model.
-
Learn enough Python
Variables, lists, loops, functions. About a week, an hour a day.
Python for AI -
Understand what a model is
What "learning from data" means, before any maths shows up.
Machine learning -
Build something small
A spam detector you can run on your own laptop. Full code included.
Projects
Some of this is genuinely hard the first time. That is not a sign you are bad at it. Read the confusing part twice and keep going.