All lessons
All 93 AI lessons
Every lesson in the AI Foundation roadmap, track by track. Pick one to see what it teaches.
AI Basics: Start Here
Never studied AI? Start here. 10 short lessons on what AI is, how machines learn and how ChatGPT works. No maths or coding needed.
- What is AI? What “artificial intelligence” really means, with everyday examples.
- AI vs ML vs Deep Learning vs GenAI How the four big words fit inside each other.
- How Machines Learn from Examples Rules or learning? Watch your first model learn.
- Types of Learning Learning with answers, without answers, and by trial and error.
- Data, Features and Labels What a model really sees when it learns.
- Training vs Using a Model How a model is built, tested and then used.
- Neural Networks in One Picture Neurons, layers and weights, without the maths.
- How ChatGPT-style Models Work Tokens and next-word guessing, step by step.
- What AI Can and Can’t Do Made-up answers, bias, and when not to trust AI.
- Your Path to AI Engineering What to learn next, and how the tracks on this site fit together.
Prompt Engineering
How to ask an AI model clearly so you get useful answers. The first skill every AI user needs.
- Writing a Clear Prompt Role, task, context and format: the four parts of a good prompt.
- Examples in Prompts Show the model what you want with zero, one or a few examples.
- Thinking Step by Step Ask the model to reason before it answers, and know when it helps.
- System Prompts and Templates Set the rules once, and reuse prompts with blanks to fill.
LLM Engineering
How real AI apps are made fast, cheap, correct and safe. The 15 ideas every AI engineer must know.
- KV Cache How a model saves its past work so it does not redo it for every word.
- Prompt Caching Reusing work when many requests start with the same text.
- Semantic Caching Reusing an old answer when a new question means the same thing.
- Speculative Decoding A small model writes a draft. A big model checks it.
- Token Budgets Deciding how many tokens each part of a request can use.
- Model Routing Sending each request to a model of the right size and price.
- Context Window vs Effective Context A model can read 200K tokens. It does not use all of them well.
- Hybrid Search (BM25 + Vectors) Searching by exact words and by meaning, together.
- Reranking A second, more careful check of search results.
- Structured Outputs Getting clean JSON data with the right fields, every time.
- Tool Calling Letting a model use functions, apps and APIs.
- Eval Sets Tests that show if a change made your AI better or worse.
- LLM-as-a-Judge Using one model to check and score another model’s answers.
- Guardrails Safety checks on what goes into a model and what comes out.
- Prompt Injection When hidden text tries to give the model orders.
RAG and AI Agents
Give a model your own documents, tools and memory. The skills behind most AI apps built today.
- Embeddings Meaning stored as a list of numbers.
- Vector Databases Storing embeddings and finding the nearest ones fast.
- Chunking Documents Cutting long documents into pieces a model can search.
- RAG Step by Step Letting a model read your documents before it answers.
- Checking RAG Quality Did it find the right text, and did it answer from it?
- What Is an AI Agent? A model that plans, uses tools and works in steps.
- The Agent Loop Think, act, look at the result, repeat.
- MCP: Model Context Protocol One standard plug for connecting tools and data to AI apps.
- Agent Memory What an agent remembers within a task and between tasks.
- Multi-Agent Systems Several agents sharing one big job.
Python & Tools
The language and tools used for almost all AI work. Every lesson has Python you can run on the page.
- Python Basics Values, variables and types. Your first lines of Python.
- Lists, Dictionaries and Loops Keeping many things together, and doing something with each one.
- Functions, Modules and Errors Reusable code, using libraries, and reading error messages.
- Files, JSON and Web APIs Reading files, the JSON format, and calling a web service.
- Packages and Virtual Environments Installing packages with pip, and keeping projects separate with venv.
- Jupyter Notebooks Notebooks where you write and run code step by step.
- NumPy Fast maths on whole lists of numbers. The base of all ML libraries.
- Pandas Loading, cleaning and exploring tables of data.
- Plotting with Matplotlib Drawing charts to see your data before you build a model.
- Calling an LLM from Python Send a prompt, read the reply, stream it, and handle errors.
Math Foundations
Only the maths you need for AI, with pictures, small numbers and code you can run. Class 10 maths is enough to start.
- Vectors Lists of numbers that describe things, and how to compare them.
- Matrices Tables of numbers, and the multiplication inside every neural network.
- Derivatives The slope that tells a model which way to move.
- Gradients Many slopes at once, and how a model walks downhill.
- Probability Chances, and how a model picks its next word.
- Statistics Average, spread and patterns in data.
System Design for Beginners
How big apps like IRCTC, UPI and Swiggy serve crores of users without breaking. No experience needed.
- How the Web Works What happens between tapping a link and seeing the page.
- APIs and HTTP How apps talk to servers with requests and responses.
- Databases: SQL and NoSQL Where apps keep their data, and how to choose a database.
- Scaling: Bigger or More Servers What to do when one server is not enough.
- Load Balancers Sharing the traffic fairly between many servers.
- Caching and CDNs Keeping ready answers close to the user to save time.
- Message Queues Doing slow work later, so the user does not wait.
- Replication and Sharding Copying data for safety and splitting it for size.
- Design a URL Shortener Your first full design, step by step, like in an interview.
- Design an AI Chat App Put it all together for a ChatGPT-style app.
Classical Machine Learning
The classic ML methods, each with an animation and code you can run.
- Linear Regression Predicting a number with a straight line.
- Gradient Descent How a model walks downhill to make fewer mistakes.
- Logistic Regression Predicting yes or no, with a probability.
- k-Nearest Neighbours Deciding by looking at the nearest examples.
- Decision Trees A flowchart of questions, learned from data.
- Support Vector Machines Drawing the widest possible gap between two groups.
- k-Means Clustering Finding groups in data without labels.
- PCA Squeezing many columns of data into a few.
- Evaluation Metrics Accuracy, precision, recall, F1 and ROC, explained simply.
- Overfitting & Underfitting Memorising answers versus really learning.
Deep Learning
Neural networks, from one neuron to modern designs.
- The Perceptron One artificial neuron.
- Neural Networks Joining neurons into layers.
- Activation Functions ReLU, sigmoid and tanh, and why we need them.
- Backpropagation How a network finds which weights caused the error.
- Optimizers SGD, momentum and Adam: smarter ways to walk downhill.
- Regularization Dropout, weight decay and early stopping.
- Convolutional Networks How computers see images.
- RNNs & LSTMs Networks with memory, for sequences like text.
Modern AI
The ideas behind today's chat and image models.
- Attention Letting every word look at every other word.
- Transformers The design behind ChatGPT and other LLMs.
- Large Language Models Next-word guessing at a huge scale.
- Fine-tuning Teaching a big model your own task.
- Diffusion Models Making images by slowly removing noise.
- Reinforcement Learning Learning by trial, error and reward.
Practice & Production
Turning what you know into real, responsible projects.
- Project Ideas Guided projects to build your portfolio.
- Deploying Models Putting a model online behind an API.
- MLOps Basics Saving versions, watching models, and training them again.
- AI Ethics & Safety Bias, fairness and using AI responsibly.