Time-boxed path

Learn Machine Learning Free in 3 Months

Train, evaluate, and improve machine learning models, and understand how neural networks and deep learning work.

107 hours· 6 free courses

1.2h/dayover 90 days

Three months of machine learning takes you from your first model to understanding how neural networks actually learn. This plan is 107 hours, about 1.2 hours a day. It assumes you can already write Python. You start with Kaggle's short, applied ML courses to build and improve models quickly, add Google's crash course for the concepts, then move into deep learning with MIT's intro and Karpathy's build-it-from-scratch series. By the end you understand the math and mechanics, not just the library calls.

The daily math

This plan is 107 hours of free courses spread across 90 days. That works out to about 1.2 hours a day, or roughly 8.3 hours a week if you prefer to batch it. Miss a day and you make it up on the weekend; the point is a pace you can actually keep.

107h

total

1.2h

per day

8.3h

per week

The plan, course by course

beginner

Kaggle Learn

Certificate

Intro to Machine Learning (Kaggle)

Train your first model. Three hours to see the whole ML loop end to end.

intermediate

Kaggle Learn

Certificate

Intermediate Machine Learning (Kaggle)

Handle real data. Four hours on the messy parts models actually face.

beginner

Google for Developers

15h

Machine Learning Crash Course

The concepts. Fifteen hours from Google on how ML works under the hood.

Machine Learning with Python

Build a project. Thirty hours applying ML with Python.

intermediate

MIT OpenCourseWare

30h

MIT Introduction to Deep Learning

Into deep learning. Thirty hours from MIT on neural networks and modern architectures.

advanced

Andrej Karpathy

25h

Neural Networks: Zero to Hero

From scratch. Twenty-five hours with Karpathy building a neural net line by line, which is where it all clicks.

Why this set, and what we left out

We paired applied courses (Kaggle, freeCodeCamp) with concept-first ones (Google, MIT) because ML clicks when you both build models and understand why they work. Karpathy's Neural Networks: Zero to Hero closes the plan because building a neural net from scratch cements everything the higher-level courses gloss over. We left the 120-hour Deep Learning Specialization for the 6-month AI engineer plan, where there is room for it.

The honest catch

You need Python before you start; this plan does not teach it. And three months gets you understanding and small projects, not production ML systems or MLOps. Deployment, scaling, and specialization come later.

Keep going

All free Machine Learning courses →See the full roadmap →Untimed learning path →All time-boxed paths →

Frequently asked questions

Can I learn machine learning in 3 months?

If you already know Python, three months at about 1.2 hours a day gets you training models and understanding deep learning fundamentals. Without Python first, add the 3-month Python plan.

Do I need to know Python already?

Yes. This plan assumes working Python. If you are new, do the 30-day or 3-month Python plan first, then come back here.

How much math is involved?

Some. The courses teach the intuition, but comfort with basic algebra and statistics helps a lot, and Karpathy's series goes into the actual calculus behind training.

Other 3 Months plans

Web Development

3 Months

Python

3 Months

Data Science

3 Months

Backend Development

3 Months