Why learn the fundamentals of AI?
A mental model before you touch a single library
Artificial Intelligence is the broad idea of building systems that perform tasks normally requiring human intelligence. Machine Learning is the subset where a system learns patterns from data instead of being explicitly programmed with rules, and Deep Learning is the subset of ML that uses layered neural networks to handle complex tasks like vision and language. Everything else — NLP, computer vision, generative AI, LLMs — is a specialization built on top of that same core.
Most people don't need a computer science degree or advanced math proofs to get productive here — they need a clear order: programming and just enough math, then core ML algorithms and how to evaluate them properly, then deep learning fundamentals, and only then the applied areas (language, vision, generative models) that map to real jobs.
This roadmap is the on-ramp to the more specialized roadmaps on this site — LLMs & RAG, Prompt & Context Engineering, and Agentic AI — all of which assume you already have the foundation covered here.
The Introduction to AI Roadmap
Work through these in order — each step builds on the last
What is AI?
Get a mental model of the field before writing any code.
Math & Programming Foundations
Just enough Python and math to understand what the algorithms are actually doing.
Data Handling
Every model is only as good as the data it learns from.
Core Machine Learning
The classical algorithms that still power a huge share of real-world AI.
Neural Networks & Deep Learning
Layered networks that learn complex patterns — the backbone of modern AI.
Natural Language Processing
Teach a system to work with human language.
Computer Vision
Teach a system to interpret images and video.
Generative AI & LLMs
Models that create — text, images, audio — rather than just classify or predict.
AI Ethics & Responsible AI
Know the failure modes and responsibilities that come with shipping AI systems.
Tools & Ecosystem
The environment and libraries you'll actually work in day to day.
Specializations & Career Paths
Pick a direction now that you can see the whole map.
GitHub Projects
Real, buildable projects to put on your own GitHub
Classic ML Predictor
Train a regression or classification model on a well-known dataset (housing prices, Titanic survival) with scikit-learn, and properly evaluate it with a train/test split.
Handwritten Digit Classifier
Build and train a small CNN on the MNIST dataset with PyTorch — a classic first deep learning project that covers layers, training loops, and evaluation.
Sentiment Analysis Tool
Use a pretrained transformer from Hugging Face to classify the sentiment of product reviews or tweets, and compare it against a simpler baseline model.
Image Classifier with Transfer Learning
Fine-tune a pretrained vision model on a small custom image dataset (plant species, pet breeds) instead of training a CNN from scratch.
Frequently Asked Questions
Common questions from people starting out with AI
What's the actual difference between AI, ML, and deep learning?
AI is the broad goal of building systems that behave intelligently. Machine learning is a way of achieving that — systems that learn patterns from data rather than following hand-written rules. Deep learning is a subset of ML that uses layered neural networks, which turned out to be especially good at tasks like vision and language.
Do I need a strong math background to start?
Not to start. You need conceptual comfort with vectors, matrices, basic statistics, and what a gradient is — not formal proofs. That's enough to understand what algorithms are doing; deeper math becomes more useful once you move into research or highly specialized model design.
Should I learn classical machine learning before deep learning?
Generally yes. Classical ML (regression, decision trees, clustering) teaches core concepts — overfitting, evaluation, feature engineering — with far less computational overhead, and those concepts carry directly into deep learning once you get there.
PyTorch or TensorFlow — which should I learn?
PyTorch has become the more common default for learners and researchers in 2026 — it's Pythonic and has strong Hugging Face ecosystem support. TensorFlow remains widely used in some production and mobile deployment contexts, so it's worth recognizing even if you learn PyTorch first.
Is it too late to get into AI in 2026?
No — demand for people who understand AI fundamentals keeps growing across healthcare, finance, and nearly every other industry, and the field keeps producing new specializations (agentic AI, LLM engineering) that are still young. The fundamentals in this roadmap don't go stale even as the applied layer moves fast.
How do I pick a specialization once I've covered the basics?
Build one small project in a couple of areas — NLP, computer vision, generative AI — and notice which one you keep wanting to go back to. From there, this site's LLMs & RAG, Prompt & Context Engineering, and Agentic AI roadmaps go deep on the generative/LLM path specifically.
Track complete
From "what is AI" to a trained model you can explain and evaluate — that's the foundation nearly every AI role is built on. From here, the LLMs & RAG, Prompt & Context Engineering, and Agentic AI roadmaps take you into the generative AI layer that's driving most of today's hiring.
Where next?
Keep exploring by domain or drill into a single skill