Curriculum atlas: maths, machine learning and AI
Seven tracks ordered from foundations to production, after an optional maths primer you can skip. Each one synthesises the standard textbooks and certifications into lessons you can manipulate, and the free tracks open with a free lab.
45 of 45 lessons
- 0.1Free labFunctions and graphs: a model is a function with knobs25 min
- 0.2Free labVectors: arrows, adding, scaling and the dot product30 min
- 0.3Free labMatrices: tables of numbers that move space30 min
- 0.4Free labDerivatives: slopes, tangents and the chain rule30 min
- 0.5Free labPartial derivatives and gradients: walking downhill30 min
- 0.6Free labProbability: chance, averages, spread and Bayes35 min
- 0.7Free labExponentials, logarithms, sigmoid and softmax30 min
Track 01
Mathematical Foundations
Eigen-space and SVD transformer
- 1.1Free labLinear transformations, eigenvectors and SVD in 3D35 min
- 1.2Lab, free previewGradients, Jacobians and the chain rule30 min
- 1.3Lab, free previewProbability, Gaussians and maximum likelihood32 min
- 1.4Lab, free previewConvexity, Lagrange multipliers and duality34 min
- 1.5Lab, free previewEntropy, cross-entropy and KL divergence30 min
Track 02
Classical Machine Learning
Gradient descent and loss landscape explorer
- 2.1Free labGradient descent, saddle points and optimizer dynamics40 min
- 2.2Lab, free previewLinear and logistic regression from first principles30 min
- 2.3Lab, free previewBias, variance and regularisation28 min
- 2.4Lab, free previewDecision trees, random forests and gradient boosting36 min
- 2.5Lab, free previewSupport vector machines and the kernel trick34 min
- 2.6Lab, free previewk-means, Gaussian mixtures and EM34 min
- 2.7Lab, free previewPCA and dimensionality reduction32 min
- 2.8Lab, free previewPrecision, recall, ROC and calibration32 min
Track 03
Deep Learning and Autograd
Neural decision manifold and backprop graph
- 3.1Free labBackpropagation and non-linear decision boundaries45 min
- 3.2Lab, free previewBuild a scalar autograd engine40 min
- 3.3Lab, free previewInitialisation, normalisation and training stability32 min
- 3.4Lab, free previewConvolutions, receptive fields and feature maps34 min
- 3.5Lab, free previewRecurrent networks, LSTMs and backprop through time36 min
Track 04
Transformers and LLMs
Multi-head attention and embedding geometry
- 4.0Free labInside an LLM: from one neuron to the next word45 min
- 4.1Lab, free previewSelf-attention geometry and scaled dot-product in 3D42 min
- 4.2Lab, free previewTokenisation and embedding spaces30 min
- 4.3Lab, free previewMulti-head attention and the transformer block38 min
- 4.4Lab, free previewPretraining, sampling and fine-tuning LLMs40 min
- 4.5Lab, free previewEmbeddings, vector search and retrieval-augmented generation38 min
- 4.6Lab, free previewAlignment: reward models, RLHF and DPO36 min
- 4.7Lab, free previewLLM agents, tool use and evaluation34 min
Track 05
Generative AI and Diffusion
Langevin score field and reverse diffusion
Drift, quantisation and latency frontier
- 6.1Lab, free previewQuantisation and efficient inference32 min
- 6.2Lab, free previewDistribution shift and drift monitoring28 min
- 6.3Lab, free previewTraining pipelines, feature stores and serving30 min
- 6.4Lab, free previewLLM serving: KV cache, batching and speculative decoding32 min
- 6.5Lab, free previewResponsible AI, fairness and generative AI security30 min