Learn the math.
Code the ideas.
Build an intuition for machine learning.
Connect the concepts, put them into code, and see how far you’ve come.
Where learning takes shape.
↳ Gradient descent over a non-convex loss surface.
A whole field.
A place to begin.
It clicks when
it all connects.
A lecture opens a question. A graph makes it tangible.
A few lines of code make it yours.
Find your next “aha.”
Follow the connections from first principles to the models you want to build.
Find your starting point 02Think it. Implement it.
Turn the equation into working Python. Test your assumptions, one function at a time.
Step into Tensor Code 03Make the learning stick.
See your progress take shape and revisit ideas before they slip away.
See the bigger pictureFollow your curiosity.
Give it a direction.
Start with the fundamentals. Go deep on a model.
Build something that makes it all click.
There’s more than one way into machine learning.
Build your foundations
The ideas underneath every model.
Understand it. Then implement it.
Think in models
Give those ideas something to learn.
Make it real
From attention to working systems.
A guide, not a fixed order. Make it yours.
Mathematical Foundations
Core mathematics underpinning machine learning — linear algebra, calculus, probability, optimization, information theory, and numerical methods.
Explore the full curriculumConcepts, curated resources, and original papers.
Beyond “I get it.”
To “I built it.”
Implement the algorithms you’re learning.
Inspect the outputs. Understand why they work.
import numpy as npdef softmax(logits, temperature=1.0): # Small change. Different confidence. z = np.array(logits) / temperature z = z - np.max(z) exp_z = np.exp(z) return exp_z / np.sum(exp_z)logits = [2.0, 1.0, 0.1, -1.0]probs = softmax(logits, temperature)What does confidence look like?
Move the temperature. Watch the distribution.
From your first NumPy array to your own attention mechanism.
Explore 146 problemsSmall steps.
A different you.
A topic understood. A problem solved. A concept remembered a month later. This is what getting better looks like.
- A living picture of what you know
- Confidence-based spaced reviews
- Study sessions, streaks, and milestones
just keep updating your weights.
A little more capable.
Every connection counts.
Which direction does the gradient point?
Curious?
Good starting point.
Where should I start?
Start with the curriculum and mark the topics you already know. Prerequisite links help you find your next step. If you’re new to ML, begin with Mathematical Foundations and the NumPy library drills.
What can I code here?
There are 146 Python problems covering mathematical foundations, ML algorithms, deep learning, paper implementations, and library practice. The full editor includes starter code, hints, visible tests, and hidden tests when you submit.
How does the review system work?
Rate your confidence when you complete a topic. That schedules a review from one day to one month later. Remember it and the interval grows; feel hazy and it shrinks. Forget it and the topic reopens for learning. Daily recall quizzes feed the same schedule.
Do I need a GPU or a local setup?
Just a browser and a Google account. Python exercises run in an isolated workspace, so you can focus on the implementation without setting up a local environment.
Can I make my own learning path?
Yes. Create an ordered collection of topics, make it your active path, and follow your progress through it. You can keep a path private, share it, or fork a public path and adapt it.
Big ideas.
Daily progress.
A little more understanding. A little more practice. Every day.
Start learning Sign in with Google. Start wherever you are.