ML Engineer: Zero to Hero
A zero-to-hero path through the catalog — every phase ends in a capstone you can put in a portfolio.
Math → classical ML in depth → deep learning → designing, deploying and operating ML systems in production.
1
Foundations0/8 lessons · capstoneThe programming and math every model in this path is built from.
Programming fundamentalsEnough data structures and complexity to reason about a training pipeline's cost.Math for MLLinear algebra, probability, and calculus — the language every model is written in.The map of MLThe three axes that place almost any ML system, before you learn any one model.2
Classical ML in Depth0/24 lessons · capstoneThe models that still win on tabular data, and the evaluation discipline that tells you whether they did.
Supervised modelsPredicting a label or a number from examples, from nearest neighbours to boosted trees.Optimization & regularizationHow models are fit, what the loss surface looks like, and how to stop them memorizing.Evaluating models honestlyHow to know if a model actually works, and which kind of failing it is when it doesn't.Unsupervised learningFinding structure with no labels — and why high dimensions make that harder than it looks.3
Deep Learning0/27 lessons · capstoneNeural networks from a single neuron to convolutional, recurrent and generative models.
Neural network fundamentalsHow a network learns at all, and why deep ones used to stop learning.Training in practiceWriting the training loop yourself, then building intuition by training networks live in the browser.VisionConvolutions, pooling, and the full pipeline from pixels to a class.Sequences & generative modelsModels for ordered data, the attention mechanism that replaced most of them, and models that generate.4
Designing ML Systems0/6 lessons · capstoneBefore any model: whether this should be ML at all, what it must achieve, and how to frame it.
Before the modelTurning a business problem into an ML problem — or deciding it isn't one.Worked system designsThree ML systems designed end to end, at the depth an interview expects.5
Training Data0/6 lessons · capstoneMost of the work, and most of the bugs, live in the data rather than the model.
Data engineeringFormats, storage, and how data moves between the services that produce and consume it.LabelsGetting labels when they are scarce, expensive or skewed toward one class.FeaturesBuilding features that help the model without quietly leaking the answer.6
Modeling & Deployment0/2 lessons · capstoneChoosing and evaluating the model offline, then getting predictions to the people who need them.
Model developmentBaselines, experiment tracking, and offline evaluation that goes beyond one aggregate number.DeploymentBatch versus online prediction, latency budgets, and making a model small enough to serve.7
ML in Production0/7 lessons · capstoneA model degrades the day it ships. Monitoring, retraining, platform, and the people it affects.
Monitoring & adaptationDetecting when the world has moved, updating the model, and proving the update helped.Infrastructure & platformThe layers under every production model, and what to build versus buy.People & responsibilityThe users, the team, and the harms an ML system can cause.