A beginner’s map for learning AI: from the basics to a first project

“I want to learn artificial intelligence, but where do I start?” This is one of the questions we hear most often from students, parents and professionals who want to change direction. The reason is easy to understand: the field is wide, the terminology is dense, and the material online is huge and unordered. Many people get lost between a course that opens with advanced mathematics and another that jumps straight to ready-made tools without any understanding. In this article we offer a realistic map for learning AI in stages – what you learn at each stage, a small project that proves you understood it, a rough timeline, and advice that keeps you away from the common mistakes.

First, decide on your goal

Before any plan, ask yourself: what do I want to do with AI? There are two main paths, and both are worthwhile:

• The professional user: a teacher, marketer, business owner or student who wants to use the tools intelligently to improve their work. This needs a good understanding of how models work, a command of prompt engineering, and knowledge of the tools’ limits and the ethics of using them – without advanced mathematics.
• The builder or developer: someone who wants to build systems, train models and work in technical roles such as machine learning engineer or data scientist. This needs programming, mathematics and statistics in more depth.

The map below is aimed mainly at the second path, but the first and last stages are useful to everyone.

The map in five stages

Stage one: programming and data

Python is the most widely used language in this field, thanks to its simplicity and its rich libraries. Learn variables, conditions, loops, functions, lists and dictionaries, then work with data files. If you are starting from zero, read why start programming with Python. After that, meet libraries such as NumPy for numerical work, pandas for analysing tables and Matplotlib for charts.

Stage project: analyse a class’s marks file: calculate the averages, find the hardest subject and plot the distribution of marks.

Stage two: just enough mathematics

Do not let mathematics become a wall that stops you from starting, and do not ignore it entirely either. You need to understand vectors, matrices and how to multiply them; the idea of a derivative and a gradient; and the basics of probability and statistics such as the mean, the standard deviation and distributions. The aim is to understand what happens inside the algorithms, not to prove theorems. Many people learn the maths alongside stage three and come back to it whenever they need a deeper understanding.

Stage three: machine learning

Here you learn how a computer learns from examples. Start with the big ideas: supervised learning (when we have correct answers to train on) and unsupervised learning (when we look for patterns without answers). Learn core algorithms such as linear regression, logistic regression, decision trees and clustering. More important than the algorithms: how to split data into training and test sets, how to measure a model’s quality, and what it means when a model memorises its training data instead of learning from it – the problem known as overfitting. The scikit-learn library is the usual tool at this stage.

Stage project: a model that predicts the price of a used car from its specification, or sorts email into important and unimportant, with a short report on its accuracy and its errors.

Stage four: deep learning

Neural networks are the foundation of the current leap in AI. Learn the idea of an artificial neuron, layers, activation functions, and how a network learns by gradually adjusting its weights to reduce error. Then meet the main families: convolutional networks for images and the transformer architecture for text. The following video from 3Blue1Brown is an excellent visual start to understanding a neural network:

Stage project: a network that recognises handwritten digits, using a well-known teaching dataset such as MNIST and a library such as PyTorch or TensorFlow.

Stage five: modern models and their applications

With a solid foundation, modern ideas become much easier: large language models, retrieval-augmented generation to make them answer from your documents, and agents that carry out multi-step tasks. Learn to use their APIs, evaluate their results, and deal with questions of privacy, bias and safety.

Stage project: an assistant that answers questions from a set of study files, naming the source of each answer.

A rough timeline

The time varies a great deal with your background and the hours you give each week. The table below is a general estimate for a learner who gives 8 to 10 hours a week and starts from zero in programming:

Common mistakes that slow learning down

• Collecting courses without finishing them: one completed course with a project beats ten you started and abandoned.
• Watching without writing code: real learning happens when you write it yourself, make mistakes and fix them. Practise more than you watch.
• Jumping to ready-made tools without understanding: tools change fast; concepts last. Anyone who understands the foundations learns a new tool easily.
• Excessive fear of mathematics: start with programming and practice, and learn the maths when you need it to understand something specific.
• Working alone: learning with a group or a teacher gives you feedback and protects you from discouragement.

How do you measure your progress?

The best measure is your project portfolio. At the end of each stage you should have a project you can explain to someone else: what was the problem? What was the data? What did you try? What was the result? What did you learn from your mistakes? Publish your projects on a platform such as GitHub and write a short description for each. That portfolio is stronger than any certificate when you apply for an internship or a job.

If the learner is a child or a teenager

The same map works for younger learners, in a gentler order. From around the age of ten, children can start with computational thinking and visual programming, then move on to Python in lower-secondary school. The focus here is on enjoyable projects: a small game, a program that answers questions, or a simple image-classification experiment with educational tools. It also matters that children learn early that AI makes mistakes, that they should check what the tools say, and that they should protect their privacy by not sharing their photos or personal data. A parent’s role here is encouragement and follow-up, not pressure to rush through the stages.

Start today

At Badr Academy Education we designed the artificial intelligence course to take you through this journey step by step, with real projects and follow-up from the instructor. If you are starting from zero in programming, browse the academy’s courses to find the right starting point. You can get a personal plan for your age, level and interests with the Find your path tool, with hand-picked resources for each stage in the learning library.