Applied Machine Learning is a 14-chapter, hands-on course for people who can program and want to understand machine learning, not just call it.
It starts from the basic terminology and the workflow of an honest experiment, then covers why models generalize, classic models on tabular data, and neural networks. From there it moves to other kinds of data, like text and images, introduces transformers, the architecture at the heart of today's large language models, and ends with a chapter on the ethics of prediction.
By the end you should be able to:
Each chapter comes with quizzes and exercises, so you know the understanding is real and the skills are your own.
The first two chapters are free to read, no signup. The founding cohort starts Nov 2: free, seats picked by hand.
Less than you probably think, and something quite specific.
If you are not sure whether that describes you, start the free lesson and find out. It is a better test than any checklist.
About 4–8 hours a week: reading the chapter, its quizzes, and an exercise or two. A few chapters run heavier, and the founding cohort will tell us the real number.
Roughly one chapter a week, over 16 weeks. Chapter 2 and Chapter 10 get two weeks each — chapters are not all the same size, and pretending otherwise just makes two weeks of the course quietly brutal. Chapters open on a schedule rather than all at once, so the cohort moves together and each chapter gets improved by the people ahead of you. Nothing closes again.
The full week-by-week schedule → Every date, what opens when, and the breaks. That page is the only place the calendar lives, so it is the one to trust.
Marking a chapter read is not the same as finishing it. The quizzes and exercises are how you, and we, know it stuck.
Uri Schonfeld wrote this course. He is a Continuing Instructor at UC Berkeley, where he teaches applied machine learning.
Uri's AI TA is an AI. It runs on a large language model; it is not a person, and it is not Uri, so it never speaks for him. You can be as confused as you actually are in front of it without it costing you anything. It can be wrong, so push back when something does not add up.
One thing it will not do is show you its instructions, and the reason is mundane: they contain the model answers to the quizzes.
The first two chapters are open to anyone, with no signup. Reading them costs nothing; you need an email address the first time you want to ask your AI TA something or have it look at an answer, and that email creates your account on the spot. The chapters after those open week by week to cohort members.
Cohort #0, the founding cohort, is small, picked by hand, and free for the whole course, in exchange for honest feedback. Later cohorts will likely be paid. Apply → Can't start then? The same application can ask for one email when the next cohort has a date.