Curriculum
Zero-to-hero paths through the catalog — every phase ends in a capstone you can put in a portfolio.
Math & ML foundations → deep learning → the full LLM application pipeline → production AI systems.
1
Software & Math Foundations0/9 lessons · capstoneYou are still a software engineer first. Don't skip this.
Where this path ends upWhat an AI engineer actually does, and how to decide whether a feature should use AI at all.Programming fundamentalsThe data structures behind every ML library.Math for MLLinear algebra, probability, and calculus — the language every model is written in.2
Classical Machine Learning0/19 lessons · capstoneBefore neural networks, understand the simpler models that still power most production ML.
Supervised learning basicsPredicting a label or a number from examples.Evaluating models honestlyHow to know if a model actually works, not just on the data it saw.Tree-based methods & optimizationDecision trees, ensembles, and the gradient descent that trains almost everything.ML systems in productionThe system-design problems unique to ML — ranking, prediction under a latency budget, and the infrastructure underneath both.3
Deep Learning0/26 lessons · capstoneNeural networks from a single neuron up to transformers.
Neural network fundamentalsHow a network learns at all.Training in practiceWriting the loop yourself, then building intuition by training networks live in the browser.ArchitecturesThe specialized shapes networks take for images, sequences, and text.Text & representationHow raw text becomes numbers a model can use.4
LLM Foundations & Data0/16 lessons · capstoneWhere AI engineering diverges from ML research: working with a pretrained model as a component.
What a foundation model isWhat it was trained on and how — which explains most of what it will and won't do for you.Talking to a modelGetting reliable behavior out of a model you didn't train.Beyond text: seeing, hearing, generatingModern AI engineering is multimodal — images, voice, and generated media are part of the job.Knowledge & data modelingStructuring the world so a model (and your app) can query it.5
Retrieval & Grounding0/7 lessons · capstoneGiving a model facts it wasn't trained on, reliably.
Retrieval-augmented generationThe open-book-exam pattern that underlies most production LLM apps.6
Orchestration & Agents0/9 lessons · capstoneChaining calls, tools, and multiple models into something that can act.
FrameworksThe plumbing that ties prompts, retrieval, and tools together.AgentsLetting a model decide what to do next, and keeping it on a leash.7
Evaluation, Ops & Production0/23 lessons · capstoneEverything that separates a demo from a system you can trust in production.
Evaluating modelsMeasuring a model that produces open-ended text, and choosing one on evidence rather than a leaderboard.Shipping safelyKnowing when your system is actually good, and catching it when it isn't.Adapting modelsMaking a model behave the way you need when prompting and retrieval run out.Serving & inference optimizationWhere latency and cost actually go, and the techniques that win them back.Architecture & the feedback loopAssembling every layer of this phase into one application that gets better from its own users.8
Production Systems Engineering0/15 lessons · capstoneAn AI engineer who can't ship a reliable service isn't done yet — the same infrastructure discipline every backend needs.
The interview frameworkThe repeatable approach every system design interview rewards — learn it once, apply it to anything.Data layerWhere your app's and your model's state actually lives.Serving infrastructureLoad, caching, and isolating failure in a live system.Engineering practiceCode quality and testing discipline apply to AI systems too — arguably more, since correctness is harder to eyeball.