🧠 AI Roadmap

Learn AI from zero — 50 free, hands-on lessons from your first line of Python to shipping an LLM app.

🌐 Open the site  ·  ▶️ Start lesson 1  ·  🗺️ Roadmap  ·  🛠️ Projects  ·  💬 Interview prep  ·  📑 Cheat sheets  ·  🎯 What's next

---

✨ Why this roadmap

Most people learning AI drown in scattered tutorials with no order and no idea what to learn next. This is one dependency-ordered path — every lesson tells you exactly what it assumes you already know, so you always know where you are and what comes next. It runs from "what even is AI?" all the way to fine-tuning models, building agents, and shipping them.

• 🆓 Free and open forever — MIT code, CC BY 4.0 content. No sign-up, paywall, or ads.
• 📖 Learn right here on GitHub — every lesson below is a plain Markdown file you can read without leaving. The website is the same content with progress tracking, search, and a visual roadmap.
• ✅ Runnable, verified code — every sample was actually executed; the output you see is the output it produces.
• 🧩 Build as you go — 30+ project ideas, each mapped to the lesson it builds on.

🚀 Quick start

You want to…	Do this
Just learn	Open the web app — sidebar, search (⌘K), progress tracking, dark/light.
Read on GitHub	Start at What AI, ML, Deep Learning and GenAI Actually Are and follow the curriculum below, top to bottom.
Run it locally	npm install → npm run dev. Build with npm run build, test with npm test.

New to all this? Do the lessons in order. Already know some? Each lesson lists its prerequisites — skip ahead whenever you already have them.

🗺️ The curriculum

50 lessons · 7 stages · ~42 hours. Seven stages, each building on the last:

flowchart LR
    S0["🧭 Orientation<br/><small>2 lessons</small>"]
    S1["🐍 Python & Data<br/><small>5 lessons</small>"]
    S2["📐 Math You Actually Need<br/><small>3 lessons</small>"]
    S3["🌳 Classical ML<br/><small>11 lessons</small>"]
    S4["🧠 Deep Learning<br/><small>9 lessons</small>"]
    S5["🤖 Generative AI & LLMs<br/><small>17 lessons</small>"]
    S6["🚀 Ship It<br/><small>3 lessons</small>"]
    S0 --> S1 --> S2 --> S3 --> S4 --> S5 --> S6
    class S0 s0;
    class S1 s1;
    class S2 s2;
    class S3 s3;
    class S4 s4;
    class S5 s5;
    class S6 s6;
    classDef s0 fill:#8b5cf622,stroke:#8b5cf6,stroke-width:2px,color:#e5e7eb;
    classDef s1 fill:#6366f122,stroke:#6366f1,stroke-width:2px,color:#e5e7eb;
    classDef s2 fill:#38bdf822,stroke:#38bdf8,stroke-width:2px,color:#e5e7eb;
    classDef s3 fill:#22d3ee22,stroke:#22d3ee,stroke-width:2px,color:#e5e7eb;
    classDef s4 fill:#2dd4bf22,stroke:#2dd4bf,stroke-width:2px,color:#e5e7eb;
    classDef s5 fill:#f59e0b22,stroke:#f59e0b,stroke-width:2px,color:#e5e7eb;
    classDef s6 fill:#fb718522,stroke:#fb7185,stroke-width:2px,color:#e5e7eb;

Loading

Click any lesson to read it right here on GitHub.

Stage	Focus	Lessons
🧭 0 · Orientation	Get your bearings before writing code	2
🐍 1 · Python & Data	The everyday tools — Python, NumPy, pandas, plots	5
📐 2 · Math You Actually Need	Just enough linear algebra, calculus & statistics	3
🌳 3 · Classical ML	How machines learn from tables — regression to ensembles	11
🧠 4 · Deep Learning	Neural networks, from one neuron to the transformer	9
🤖 5 · Generative AI & LLMs	How LLMs work, and how to build with them	17
🚀 6 · Ship It	Take a model off your laptop and ship it	3

🧭 Stage 0 · Orientation  —  2 lessons

#	Lesson	Time	Level	What you'll learn
00	What AI, ML, Deep Learning and GenAI Actually Are	20m	beginner	The four words everyone mixes up, sorted out once, with a mental model you can keep.
01	How to Learn AI Without Burning Out	15m	beginner	A realistic schedule, the order to learn things in, and the three traps that stop most beginners.

🐍 Stage 1 · Python & Data  —  5 lessons

#	Lesson	Time	Level	What you'll learn
02	Python Basics	60m	beginner	Variables, lists, dicts, loops, functions, imports, and how to read the error messages you'll see constantly.
03	NumPy	45m	beginner	Why arrays beat lists for numeric work, plus shape, dtype, indexing, broadcasting, and vectorised math.
04	Pandas	50m	beginner	Series and DataFrames, loading CSVs, selecting and filtering rows, grouping, and handling missing values.
05	Data Visualization	40m	beginner	Choosing the right chart type, matplotlib basics, plotting straight from pandas, and reading what a histogram tells you.
06	Real Datasets	45m	beginner	Where to find datasets, loading messy CSVs, fixing types and dates, removing duplicates, spotting outliers, and a reusable cleaning checklist.

📐 Stage 2 · Math You Actually Need  —  3 lessons

#	Lesson	Time	Level	What you'll learn
07	Linear Algebra	50m	beginner	Vectors and matrices as the data structures behind every model, dot products as weighted sums, and matrix multiplication as batch prediction.
08	Calculus	50m	intermediate	Derivatives as slope, the chain rule, and gradient descent implemented by hand in NumPy — how a model actually learns.
09	Probability & Statistics	50m	beginner	Distributions, mean and variance, conditional probability, Bayes' theorem, sampling, and what a p-value actually means.

🌳 Stage 3 · Classical ML  —  11 lessons

#	Lesson	Time	Level	What you'll learn
10	ML Fundamentals	45m	beginner	Supervised vs unsupervised learning, the train/validation/test split, and overfitting vs underfitting through the bias-variance tradeoff.
11	Regression	50m	beginner	Linear regression from the normal equation to scikit-learn, the MSE/MAE/R² metrics, and Ridge/Lasso regularisation to fight overfitting.
12	Classification	50m	beginner	Logistic regression and the sigmoid, decision boundaries, k-nearest neighbours, and strategies for more than two classes.
13	Model Evaluation	45m	intermediate	The confusion matrix, precision, recall, F1, ROC-AUC, cross-validation, and why accuracy alone can make a worthless model look great.
14	Feature Engineering	45m	intermediate	Scaling, encoding categoricals, dates, binning, interaction terms, and data leakage — raising a model's score with features, not a new algorithm.
15	Clustering & PCA	45m	intermediate	k-means clustering, choosing k with the elbow method, hierarchical clustering, and PCA as compression with explained variance.
16	Trees & Ensembles	50m	intermediate	Decision trees and how splits are chosen, random forests as bagging, gradient boosting, and feature importance.
17	First ML Project	90m	intermediate	A full guided ML pipeline end to end: problem framing, EDA, cleaning, features, baseline, iteration, evaluation, and writing up results.
37	Explainable AI — Opening the Black Box	45m	intermediate	Why a model made a prediction — feature importance, permutation importance, and reading a model you can't see inside.
38	Time-Series Forecasting	50m	intermediate	Predicting what happens next — why time data breaks normal ML, lag features, a proper time-aware split, and honest baselines.
45	Recommender Systems	45m	intermediate	The engine behind every 'you might also like' — content-based vs collaborative filtering, similarity, and the cold-start problem.

🧠 Stage 4 · Deep Learning  —  9 lessons

#	Lesson	Time	Level	What you'll learn
18	Neural Networks	55m	intermediate	What one neuron computes, why nonlinear activations are non-negotiable, how depth builds representations, and the universal approximation intuition.
19	Backprop & Training	60m	intermediate	How the chain rule turns one output error into a gradient for every weight in a network, and how learning rate, epochs and batches shape training.
20	PyTorch	60m	intermediate	Tensors, autograd, nn.Module, optimisers, and the canonical training loop — the framework that automates the backprop you just wrote by hand.
21	CNNs & Vision	55m	intermediate	Convolution as a learned filter, stride, padding and pooling, how a CNN's shapes flow layer to layer, and transfer learning with a pretrained backbone.
22	Sequence Models	50m	advanced	Why order matters, how RNNs process sequences step by step, the vanishing gradient problem, LSTM/GRU, and why attention replaced them.
23	Transformers	70m	advanced	Attention as a learned lookup over query, key and value, self-attention and multi-head attention, positional encoding, and the encoder/decoder split.
39	Reinforcement Learning	55m	advanced	Learning from reward instead of labels — agents, states, actions, rewards, and Q-learning taught by making an agent solve a tiny grid world.
46	Object Detection & Segmentation	50m	advanced	Beyond 'what's in this image?' to 'what's where?' — bounding boxes, IoU, non-max suppression, and how YOLO detects in real time.
47	Generative Adversarial Networks (GANs)	50m	advanced	Two networks in a duel — a generator faking data and a discriminator catching fakes — and how that adversarial game learns to create.

🤖 Stage 5 · Generative AI & LLMs  —  17 lessons

#	Lesson	Time	Level	What you'll learn
24	How LLMs Work	55m	intermediate	Tokenization, next-token prediction, pretraining vs post-training, context windows, temperature and sampling, and why models hallucinate.
25	Prompt Engineering	45m	beginner	Clear instructions, few-shot examples, chain-of-thought, structured output, system prompts, and fixing a failing prompt in documented iterations.
26	Embeddings	45m	intermediate	Text as vectors, cosine similarity, embedding models, vector databases, and chunking strategy, with a real semantic search built in NumPy.
27	Retrieval-Augmented Generation (RAG)	60m	intermediate	Why retrieval beats stuffing the context window, the ingest-chunk-embed-retrieve-generate pipeline, chunk sizing, and common RAG failure modes.
28	Fine-tuning	60m	advanced	When fine-tuning beats RAG or prompting, full fine-tuning vs LoRA/PEFT, dataset preparation, and evaluating the result, with a real LoRA parameter-count demo.
29	AI Agents	60m	advanced	Tool use, the reason-act loop, planning, memory, multi-step failure modes, and cost control, with a real non-LLM demo of the loop mechanics.
30	Evals & Guardrails	50m	advanced	Why manual spot-checking doesn't scale, building an eval set, LLM-as-judge and its biases, regression testing, and input/output guardrails.
34	Run Open LLMs Locally with Ollama	45m	intermediate	Run real language models on your own machine — private, free, offline — and call them from Python like an API.
35	Build an Agent Harness	60m	advanced	The scaffolding that turns a raw language model into an agent: the tool loop, parsing, dispatch, and history — built from scratch.
36	Loop Engineering	55m	advanced	The naive agent loop breaks in a dozen ways. The engineering that makes it reliable: budgets, loop detection, context control, and error recovery.
40	Diffusion Models & Image Generation	55m	advanced	How Stable-Diffusion-style models turn noise into images — the forward noising process, learning to denoise, and why it works.
41	Vector Databases	45m	intermediate	Where embeddings live at scale — similarity search, why brute force stops scaling, and how approximate nearest neighbors makes it fast.
42	Structured Outputs & Function Calling	45m	intermediate	Getting reliable JSON out of an LLM — schemas, validation, function/tool calling, and what to do when the model returns something wrong.
43	Model Context Protocol (MCP)	45m	advanced	The USB-C of AI tools — a standard protocol that lets any agent connect to any tool or data source without custom glue for each one.
44	AI Ethics & Responsible AI	45m	intermediate	Building AI that doesn't cause harm — bias and fairness, privacy, transparency, misuse, and the responsibility that comes with shipping models.
48	Speech & Audio AI	45m	intermediate	How machines hear — turning sound into spectrograms, speech-to-text with Whisper, and text-to-speech, from waveform to model input.
49	Multimodal AI (CLIP & Vision-Language Models)	50m	advanced	One model, many senses — how CLIP puts images and text in the same space, enabling zero-shot classification and image search.

🚀 Stage 6 · Ship It  —  3 lessons

#	Lesson	Time	Level	What you'll learn
31	MLOps Basics	45m	intermediate	Experiment tracking, model and data versioning, reproducibility, and drift monitoring — what keeps a shipped model trustworthy after it leaves your notebook.
32	Deploying a Model	55m	intermediate	Wrapping a trained model behind a validated FastAPI endpoint, measuring latency and batching, Dockerizing it, and free hosting options.
33	Portfolio and Career	40m	beginner	What makes a project worth showing, writing a README that gets read, the real AI job families, and how to keep learning after this roadmap.

Prefer a table of everything at a glance? See CURRICULUM.md.

🛠️ Build projects

You learn AI by making things. The Projects page has 30+ ideas across four tiers, each mapped to a lesson:

• 🟦 Beginner — Titanic predictor, a stats CLI, an MNIST digit recognizer, a tic-tac-toe AI.
• 🟪 Intermediate — semantic search, transfer-learning image classifier, an end-to-end churn model.
• 🟥 Advanced / GenAI — a RAG chatbot over your docs, LoRA fine-tuning, an AI agent, reproduce nanoGPT.
• 🟨 Capstone — a full-stack AI product, a Kaggle competition, reproduce a paper, an open-source contribution.

💬 Interview prep & cheat sheets

Two extras to get you job-ready and keep you fast:

• Interview questions — the ML, deep-learning, LLM, and system-design questions that actually come up, each with a clear answer and a link to the lesson that teaches it.
• Cheat sheets — one-page quick reference for Python, NumPy, pandas, scikit-learn, PyTorch, prompting, and the core ML concepts. Bookmark it.

🎯 After the roadmap

Finished? The What's next page is a curated guide to going further — specializing (computer vision, RL, diffusion, LLMs from scratch), practicing on Kaggle, free courses & books, communities, MLOps, and turning it all into a career. Every resource is free.

💡 How to get the most out of it

• Run every code sample. Don't just read it — type it, break it, change the numbers. That's where the learning is.
• Do the "Build this" exercise at the end of each lesson before moving on.
• Ship 3 projects as you go (a data project, a deep-learning project, an LLM app). A public repo beats any certificate.
• A few focused hours a week beats cramming — the whole path is ~42 hours of reading.

🧑‍💻 Built with

Astro static site · TypeScript · Markdown content · Shiki highlighting · deployed to GitHub Pages by CI, which validates content and runs 100+ tests on every push. The lessons are the product; the site is a nice reader over them.

🤝 Contributing

Spotted a bug or a clearer explanation? Issues and PRs are welcome.

• Lessons live in content/lessons/ as portable Markdown — no site-specific syntax, so they read cleanly on GitHub too.
• Run npm run validate -- --strict and npm test before opening a PR; CI runs both.
• Lesson order and prerequisites are defined in content/roadmap.json.

📄 License

Code is MIT. Lesson content is CC BY 4.0 — use it, remix it, just credit the source.

---

_{Built to be the roadmap I wish I'd had. ⭐ it if it helps — and start learning →}