VITON13School

VITON13 SCHOOL · Free course

Generative AI for Beginners

21 lessons from your first API call to an app you put online. Read, build, check yourself, move on.

Course by , Founder of VITON13 SCHOOL
  • 21 lessons
  • Python practice
  • Any model provider
  • Free, no sign-up
Start lesson 1See the programme

How to take this course

I rebuilt this course for one kind of student: someone who wants to make things with AI, not collect certificates. Every lesson ends with a task you can check yourself, and every task adds a piece to one project, Study Buddy. By lesson 21 it explains topics, quizzes you, searches your own notes and runs online.

Take one lesson a day. Read it once, type the example by hand instead of pasting it, then do the practice before you open the quiz. If the quiz catches you out, reread that section; do not move on with a gap.

Use whichever model you can reach: a paid API, a router such as OpenRouter or a model on your own laptop through Ollama. The code reads three settings from your environment, so switching providers never means rewriting a lesson.

Keep a short log as you go: what you asked the model, what came back, what you changed. That log becomes your evaluation set in lesson 15, and it is the fastest way to see how much you have learned.

Vitalii TarasovFounder of VITON13 SCHOOL

How this course is built

  • Any providerEvery example runs on any OpenAI-compatible endpoint, from a paid API to a model on your own laptop.
  • Short lessonsOne idea per paragraph, 30 to 60 minutes a lesson, and setup notes cut to what you need.
  • Checkable practiceEvery task says what “done” looks like, so you know when to move on.
  • Quiz and mistakesThree questions with explanations and the mistakes beginners make most, in every lesson.
  • One projectStudy Buddy grows lesson by lesson and goes online in the capstone, lesson 21.
  • Current topicsTools through MCP, prompt injection, evaluations and agents are treated as everyday work, not extras.

Course project

The project: Study Buddy

One small app that grows with you. It starts as a single API call in lesson 1 and finishes as a study assistant with chat, search over your notes, a tool, tests and a safety checklist.

  1. Lesson 1First call to a model
  2. Lesson 8A chat with memory
  3. Lesson 12A tool it can call
  4. Lesson 16Answers from your notes
  5. Lesson 21Online, tested, safe

The programme

Part 1

Foundations

How models work, how to choose one, how to use it responsibly and how to ask it well.

  1. 01Set up your AI workbenchBefore you can build anything, you need a clean Python project that can talk to a language model.35 min
  2. 02What generative AI is and how LLMs workA language model does one thing extremely well: it guesses the next piece of text.30 min
  3. 03Choosing an LLM for your taskThere is no best model, only the best fit for a task, a budget and a response time you can accept.40 min
  4. 04Using generative AI responsiblyAn AI feature can mislead, exclude or expose people without anyone intending it.35 min
  5. 05Prompt engineering fundamentalsA prompt is the main way you program a language model, and small changes in it can change the result a lot.40 min
  6. 06Advanced prompting techniquesClear instructions solve most problems, but some tasks need more: a pattern to copy, room to reason, a second look or a pipeline of small steps.45 min

Part 2

Build

Text, chat, search, images, low-code and tools: the six kinds of app most products are made of.

  1. 07Building a text generation appA prompt in a chat window is an experiment; a prompt inside a program is a product.50 min
  2. 08Building chat applicationsA chat app looks like magic memory, but the model forgets everything between calls.45 min
  3. 09Semantic search with embeddingsKeyword search finds the words you typed; semantic search finds what you meant.40 min
  4. 10Building image generation appsText models write; image models draw from a description.40 min
  5. 11Low-code AI apps and automationsNot every AI feature needs a Python project.35 min
  6. 12Function calling and toolsA model on its own can only write text.50 min

Part 3

Make it trustworthy

Interfaces people trust, security against prompt injection, evaluations and answers grounded in your data.

  1. 13Designing UX for AI applicationsAn AI feature can be technically impressive and still fail because people trust it too much, too little or not at all.40 min
  2. 14Securing AI applicationsAn LLM app can be attacked with nothing but words.50 min
  3. 15The AI app lifecycle and evalsA prompt that worked in three test chats can fail on the fourth user, and a model update can change answers overnight.45 min
  4. 16RAG and vector databasesA model knows nothing about your lecture notes, and pasting all of them into every prompt is slow, costly and unreliable.50 min

Part 4

Go further

Open-weight and small models, agents, fine-tuning, and the capstone you put online.

  1. 17Open-weight models and running them locallySome model makers publish the trained weights, so you can download a model and run it on your own computer or server.45 min
  2. 18AI agents: loops, tools and guardrailsA chatbot answers one message; an agent works towards a goal over several steps, choosing tools as it goes.50 min
  3. 19Fine-tuning: when and howFine-tuning trains an existing model a little further on your own examples, so it picks up a format, tone or narrow skill without long instructions.50 min
  4. 20Small language modelsNot every task needs the biggest model.40 min
  5. 21Capstone: ship your Study BuddyOver twenty lessons you built the parts of an AI study assistant one at a time.60 min

Questions

Is the course really free?

Yes. All 21 lessons, examples, tasks and quizzes are open on this page, with no account needed. Only the model provider you choose may charge for API calls, and lesson 1 shows free and local options.

Do I need to know how to program?

Basic Python helps: variables, functions, lists and dictionaries. If you are new to it, take the free Python course at VITON13 SCHOOL first; it covers everything these lessons use. Open the Python course.

Which AI model do I need?

Any model behind an OpenAI-compatible endpoint: a commercial API, a router such as OpenRouter, or an open-weight model running locally in Ollama. Lesson 3 shows how to compare them on your own tasks.

Who wrote this course?

Vitalii Tarasov, founder of VITON13 SCHOOL. The lessons, tasks, quizzes and the Study Buddy project are the school's own and draw in part on Microsoft's open Generative AI for Beginners course, published under the MIT licence.

How long does it take?

About 15 hours of reading and practice. At one lesson a day you finish in three weeks.

Course by Vitalii Tarasov, VITON13 SCHOOL. Draws in part on Microsoft's Generative AI for Beginners, MIT licence, Copyright (c) Microsoft Corporation. microsoft/generative-ai-for-beginners