Machine Learning
Models that learn from examples: the classical supervised and unsupervised toolkit end to end, then neural networks — from one neuron to backpropagation, optimisers and the training clinic.
This domain is where models stop being formulas you apply and start being systems that learn. Two live pillars carry the journey. Traditional Machine Learning (P5) covers the classical toolkit end to end — regression to clustering, each method fitted, inspected and pushed until it breaks, with a whole track on evaluation so you learn early how not to fool yourself. Deep Learning & Neural Networks (P6) then rebuilds the idea of a model from a single neuron: forward passes, loss, backpropagation and optimisers, closing with a training-dynamics clinic on real runs.
The order matters. The instincts you earn on classical models — what fitting means, why overfitting happens, how to read an evaluation honestly — are exactly the vocabulary neural networks reuse at scale, and both pillars keep every concept attached to a lab you can poke. Finish the domain and you stand at the door of the AI Systems domain, where the Transformer is assembled from the pieces you now own.
P5 · Traditional Machine Learning
supervised + unsupervised end to end
14 live modules · 3 tracks
P6 · Deep Learning & Neural Networks
from the neuron to modern architectures
14 live modules · 2 tracks
P7 · Time Series & Forecasting
decomposition, classical + ML forecasting, anomaly detection
9 live modules · 2 tracks
P8 · Causal Inference & Experimentation
A/B at scale, uplift, quasi-experiments
8 live modules · 2 tracks
P9 · Reinforcement Learning
bandits → MDPs → policy methods; bridges to RLHF
7 live modules · 2 tracks