AI & ML for Beginners
A free, hands-on AI course for beginners. Start with prompt engineering — how LLMs read prompts, clear instructions, roles, few-shot examples, structured output, reasoning, iteration, prompting for code and AI safety — then build with LLM APIs in Python or JavaScript: your first API call, a chat app with memory, embeddings, semantic search and RAG. Interactive challenges in every lesson; modules on AI for coders, ML fundamentals and agents follow.
Start Learning →What Is This Track?
AI & ML for Beginners is a free course for anyone who wants to learn AI by doing, not just reading about it. It starts with the skill you will use every day: prompt engineering — writing instructions that get accurate, useful answers from AI assistants such as ChatGPT, Claude and Gemini. Then it moves from using AI to building with it: Module 4 shows you how to call AI models from your own Python or JavaScript programs and build chat, search and question-answering features. Modules on coding with AI, how machine learning works and simple agents are on the way.
Every lesson includes interactive mini-challenges — fix a bad prompt, spot what is missing, put the parts of a prompt in order, or predict what a model will do. They run right in your browser, with no account and no API key needed. Module 4 also includes complete programs in Python and JavaScript that you can run on your own computer.
Like every PrograMinds course, this track is completely free. Read the lessons in order, try each challenge, and your solved challenges are remembered in your browser.
Who It Is For
- Complete beginners who use AI chat tools and want consistently better answers.
- Students who want to learn AI for free before choosing a specialisation.
- Developers who want practical prompting skills for coding, then a path into building AI features.
- Anyone who has to evaluate AI output at work and wants to understand its limits.
What You Will Be Able to Do
Write Clear Prompts
Use task, context, format and constraints to get the answer you actually need the first time.
Get Structured Output
Ask for JSON, tables and tagged sections that people and programs can rely on.
Fix Bad Answers
Diagnose why an answer missed and improve the prompt one change at a time.
Prompt for Code
Write debugging, refactoring and testing prompts that coding assistants can act on.
Use AI Safely
Recognise hallucinations, verify facts and understand prompt injection.
Build AI Features
Call models from code and build chat, semantic search and RAG in Python or JavaScript.
Module 1: Prompting Skills
Nine lessons that take you from “what is an LLM?” to safe, systematic prompting. Follow them in order — each builds on the one before.
Module 4: Building with LLM APIs
Five lessons that take you from prompting in a chat window to building software. Every example is shown in both Python and JavaScript (Node.js), and the embedding lessons run free models on your own computer. Module 1 is the only prerequisite — basic programming knowledge in either language helps.
Want to see where these skills lead first? The AI in Apps guide shows the six patterns real products use to integrate AI models — chatbots, summarising and extracting, smart search, RAG, classification and agents — each with a working example in Python and JavaScript.
AI in Apps: How to Integrate AI Models into Real Applications
Coming Next: Modules 2, 3 & 5
These modules are being written now. Here is what each one will cover, so you can see where the track is heading.
AI for Coders
Use AI assistants as a pair programmer you supervise: reading unfamiliar code, finding bugs, refactoring safely and writing tests — while staying in charge of correctness.
Reading Unfamiliar Code with an AI Assistant
Ask an assistant to explain a file, trace a function call or map a codebase. Learn which questions give accurate answers and how to check them against the code.
- Ask for explanations at the right level of detail
- Trace how data flows through a function
- Verify an explanation against the actual code
Debugging with AI: From Error Message to Root Cause
Give the assistant the error, the code and what you expected, so it can reason about causes instead of guessing. Then confirm the fix with a reproduction.
- Package a bug report an AI can work with
- Tell symptom fixes apart from root-cause fixes
- Confirm a fix with a minimal reproduction
Refactoring Safely with AI
Improve structure without changing behaviour: small steps, explicit constraints and tests before and after every change the assistant suggests.
- Write refactoring prompts with clear "do not change" rules
- Review a diff for accidental behaviour changes
- Use tests as a safety net for each step
Writing and Reviewing Tests with AI
Generate unit tests that check real behaviour, including edge cases, and spot weak tests that only repeat the implementation.
- Ask for edge-case and failure-path tests
- Recognise tests that would pass even if the code were wrong
- Turn bug reports into regression tests
Reviewing AI-Generated Code
A checklist for code you did not write: correctness, security, performance, licensing and whether it fits your project's conventions.
- Apply a repeatable review checklist
- Spot common security mistakes in generated code
- Decide when to rewrite rather than patch
ML Fundamentals
The core ideas behind machine learning — data, features, training, testing, overfitting and evaluation — with plain-Python examples you can run in the PrograMinds compiler sandbox.
What Machine Learning Is (and Isn't)
Learning rules from examples instead of writing them by hand: where ML works well, where ordinary code is better, and how LLMs fit in.
- Explain supervised vs unsupervised learning
- Identify problems suited to ML
- Relate classic ML to large language models
Data and Features
Turn raw records into numbers a model can learn from, and see how missing values, bias and leakage quietly ruin results.
- Choose and encode useful features
- Handle missing and messy values
- Recognise data leakage and sampling bias
Training vs Test Data
Why a model must be judged on data it has never seen, and how to split a dataset fairly — illustrated with a nearest-neighbour classifier in plain Python.
- Split data into training and test sets
- Implement k-nearest neighbours without libraries
- Explain why test data must stay unseen
Overfitting and Underfitting
A model that memorises its training data fails on new data. Learn to spot both failure modes and the usual remedies.
- Recognise overfitting from training vs test scores
- Choose simpler models or more data as remedies
- Understand the bias–variance trade-off intuitively
Evaluating a Model: Accuracy, Precision and Recall
Why 95% accuracy can be useless, and how a confusion matrix, precision and recall show what a model really gets wrong.
- Build a confusion matrix by hand
- Calculate precision and recall in Python
- Pick the right metric for the problem
Agents & Tools
Let a model take actions: describe tools it can call, run the agent loop safely, and finish with a mini agent project that combines everything in the track.
Function Calling: Giving a Model Tools
Describe functions with a JSON schema, let the model request a call, run it in your code, and send the result back.
- Write clear tool names, descriptions and schemas
- Execute tool calls and return results
- Report tool errors back to the model
The Agent Loop
An agent is a loop: call the model, run the tools it asks for, repeat until it answers. Learn the loop, its stop conditions and its limits.
- Implement a basic agent loop
- Handle parallel tool calls
- Add round limits and timeouts
Designing Good Tools
Agents are only as good as their tools: small, well-described, safe-by-default functions with clear errors.
- Scope tools narrowly
- Write descriptions the model can act on
- Require confirmation for risky actions
Mini Project: A Research Assistant Agent
Build an agent that searches your notes, does calculations and writes a short report with sources.
- Combine retrieval and tools in one agent
- Structure the final report
- Test the agent on realistic questions
Safety, Limits and Evaluation for Agents
Prompt injection through tool results, runaway costs, and how to test an agent before trusting it with real work.
- Defend against injection in tool output
- Set budgets and stop conditions
- Evaluate agents with repeatable test cases
Module 3’s machine learning examples are written in plain Python, so you will be able to run them in the PrograMinds online compiler (compiler.programinds.com) without installing anything.
Frequently Asked Questions
Yes. Every lesson and challenge is free, with no sign-up. The challenges run in your browser and do not need an AI account or API key.
Not for Module 1, which is for everyone. Module 4 is about building with AI, so it uses code: every example is shown in both Python and JavaScript, and knowing the basics of either language is enough.
Only to run some examples on your own computer. Lessons 12 and 13 use free embedding models that run locally. Lessons 10, 11 and 14 call a hosted AI model, which needs an API key from a provider with pay-as-you-go pricing; each example costs a fraction of a cent. The in-browser challenges never need a key.
Any modern assistant works — ChatGPT, Claude, Gemini or similar. The principles in this course apply to all large language models; try each technique in the tool you prefer.
The AI Tools & Libraries Guide is a reference to the developer ecosystem (SDKs, LangChain, vector databases and more). This track is a step-by-step course that starts with prompting for beginners and builds towards making your own AI applications.
They are in progress and will be added as lessons are completed. The outlines above show exactly what each module will teach.
Key Takeaways & Summary
- A free, beginner-friendly AI course that starts with prompt engineering.
- Module 1 covers 9 lessons, from how LLMs read prompts to AI safety.
- Module 4 covers 5 lessons on building with LLM APIs: API calls, chat memory, embeddings, semantic search and RAG, in Python and JavaScript.
- Interactive challenges in every lesson run in your browser — no API key needed.
- Modules on AI for coders, ML fundamentals and agents are coming next.
Begin with Lesson 1, “What Is an LLM?”, and try every challenge as you go — practising is the fastest way to learn prompting.