ML Mastery

Machine learning, built from the ground up — the math included, not assumed.

Ascent is a self-paced path from your first vector to shipping original ML work. Every idea is taught in lockstep with the math underneath it, broken into short chapters, and tracked as you climb.

Start learning

24 tracks · 440+ chapters · 8 phases

The full climb, at a glance

Foundations → mastery

Phase I Foundations

01Vectorsthe language of data 02Matrices & spacestransforming that data 03Least squares & regressionyour first model 04Gradient descenthow models actually learn

Phase II How models learn

05Calculus you needno more, no less 06Linear classifiersdrawing the boundary 07Optimizersdescending faster, safer

Phase III Uncertainty

08Probability foundationsreasoning under doubt 09The Gaussian & estimationthe shape of noise 10Probabilistic modelslearning what a class looks like

Phase IV Classical ML

11Margins & SVMsthe widest safe boundary 12Trees & ensemblessplitting your way to an answer 13Unsupervised learningstructure with no labels

Phase V Neural nets

14Neurons & backprophow a network actually learns 15Training neural netsmaking it learn well 16From scratch to PyTorchthe same ideas, a real framework

Phase VI Vision & sequences

17Convolutional networksmodels that see 18Sequence modelsmodels that remember order

Phase VII Transformers

19Attention & the transformerthe architecture behind it all 20Language modelsattention, scaled up 21Generative modelslearning to create, not just label

Phase VIII Research & mastery

22Bayesian & probabilistic MLbelief, updated properly 23Reinforcement learninglearning from consequences 24Shipping & original workyour own project, built and shipped