01Available to explore
Math
Understand the language behind a prediction.
Explore functions, vectors, change, uncertainty and optimization.
Explore MathA learning path for curious beginners
Understand the ideas.
Then build models.
Connect the mathematics, Python and practical decisions behind machine learning. Start from the foundations or return to your next useful step.
Try it · about 60 seconds · nothing to install
Price a flat nobody has sold yet.
Machine learning lets a computer learn from past examples, so it can predict things nobody has told it yet: the price of a house, whether a customer will leave, whether a payment is fraud. You are about to do it yourself, in a minute, with eight house sales.
These are eight real-looking sales. Tilt and lift the line until it passes close to the dots. Then press Let it learn and watch the computer do it for you.
Average miss₹16.3 lakh
Your line is off by ₹16.3 lakh on average. Quoting the average every time is off by ₹21.6 lakh.
A 75 sq m flat → ₹61.6 lakh
What you will learn
Supervised machine learning, end to end.
Supervised learning means learning from examples whose answers are known, such as house prices or customers who left. It is the part of machine learning behind most real products, and the best place to build solid foundations.
What you will learn
Supervised machine learning, end to end.
Supervised learning means learning from examples whose answers are known, such as house prices or customers who left. It is the part of machine learning behind most real products, and the best place to build solid foundations.
- 01
Explain why a model works
Read the maths behind a prediction (functions, derivatives, probability, optimization) and say what the model is doing, not only which function to call.
- 02
Write the Python for it
From your first line of Python to NumPy, pandas, scikit-learn and your own small classes, with every code cell runnable in the browser.
- 03
Prepare data without fooling yourself
Split first, build a baseline, put preparation inside a pipeline and avoid leakage.
- 04
Build and compare supervised models
Regression and classification: linear and logistic regression, k-nearest neighbours, naive Bayes, decision trees, random forests, gradient boosting, SVMs and a first neural network.
- 05
Judge the result honestly
Choose a metric that matches the decision, use validation and a locked test set, and test whether a gain is real or luck.
- 06
Take a problem from question to a shipped model
Frame it, build it, evaluate it, package it and see how it is deployed, practised on House Price, Customer Churn and Titanic.
How the course gets you there
- Math: algebra, calculus, linear algebra, probability, statistics, optimization and 12 bridges to ML, including hypothesis tests.
- Python: six short beginner lessons, then Essentials, NumPy, pandas, visualization, data preparation and scikit-learn.
- Algorithms: 13 supervised paths with playbooks, a model-choice guide and an end-to-end workflow.
- Projects: three guided Kaggle-style projects on real public data.
Not on the roadmap yet
So you know exactly what you are signing up for. Everything else you might expect is under Coming next, below.
- Reinforcement learning
- Time-series forecasting
- Running models at scale: monitoring, retraining and MLOps beyond a first deployment
Pay once. Keep it. Updates included.
$39one-time introductory price
The first lessons are free, so you can judge the teaching before you pay anything.
Buy full access · $39 See everything included →
14-day refund. No subscription. One key works on up to 3 of your own devices.
One connected curriculum
Four sections. One learning path.
Math gives you the language. Python makes it executable. Algorithms connect the ideas. Projects bring them together. Explore freely; the order is recommended.
02Available to explore
Python
Turn an idea into code you can explain.
Build skills with Python, arrays, tables, data preparation and scikit-learn.
Explore Python03Available to explore
Algorithms
Understand how models predict, learn and fail.
Explore regression, classification and evaluation through existing learning paths.
Explore Algorithms04Full studio coming soon
Projects
Bring data, models and decisions together.
The full Project Studio and capstones are planned. Review what is available today.
View project availabilityComing next
Where this path leads.
Planned modules, in the order that makes sense: natural language processing and computer vision both rest on deep learning, so it comes first. They are plans, not dates, and none of them is available yet. Until then, the lessons listed under each one are the best preparation.
- 01
Unsupervised learningComing soon
Finding structure when there are no answers: grouping similar customers, spotting unusual transactions, squeezing many columns into a few.
You will be able to: Cluster with k-means, reduce dimensions with PCA, and judge whether the groups mean anything.
Start preparing: Distance and norms · SVD · k-nearest neighbours
- 02
Deep learning foundationsComing soon
Neural networks with many layers, trained by backpropagation, built and debugged in code.
You will be able to: Train a deeper network, read its learning curves and fix one that will not learn.
Start preparing: Gradient descent · Neural networks (MLP) · Optimization
- 03
Natural language processingComing soon
Working with text: turning words into numbers, classifying messages, searching by meaning, and how language models work.
You will be able to: Build a text classifier and explain what a language model is predicting.
Start preparing: Naive Bayes · Logistic regression · Information and entropy
- 04
Computer visionComing soon
Learning from images: how a picture becomes numbers, convolutional networks, and checking what a model looks at.
You will be able to: Train an image classifier and explain why it got a picture wrong.
Start preparing: Linear algebra · Neural networks (MLP)
At a glance
Who it is for, what you need and what it costs.
- Who it is for
- Beginners who want to understand why models work, and developers who want the maths behind the libraries.
- Prerequisites
- School arithmetic. Python is taught from zero, so no programming is assumed.
- Time
- Self-paced, in short lessons. Python from zero is six lessons of about 13 minutes each.
- Cost
- The first lessons of Math, Python and Algorithms plus one project are free. The full program is $39 one time, with updates included.
Questions, answered.
Is there a free machine learning course here?
Yes, the start is free: the first Math lessons, three Python-from-zero lessons, the linear regression introduction through gradient descent, two guides and the Titanic project. The full program is a one-time $39.
Do I need to know Python or maths first?
No. The Python lessons start from your first line of code, and the maths starts with variables and equations.
How is this different from a library-only course?
Every model is taught with the mathematics behind it and runnable Python, so you can say why a prediction was made, not only which function to call.
Does it cover deep learning, NLP or computer vision?
Not yet. It covers supervised machine learning end to end. Unsupervised learning, deep learning, NLP and computer vision are planned and listed on the home page.
Can I run the code without installing anything?
Yes. Code cells run in your browser, with nothing to install.
Our teaching direction
Learn why, then learn how.
The shared lesson structure we are developing connects a real problem to an explanation you can apply.
- Problem
- Intuition
- Experiment
- Mathematics
- Python
- Application
- Self-check
Choose the help you need.
New to the subject? Begin with variables and equations. Prefer a familiar prediction problem? Try the housing introduction.
Not sure where machine learning fits? Read the plain-language guide, or keep the Python cheat sheet open while you work. Your earlier links and records remain available.