complete guide

AI Learning Roadmap

from zero → expert · tap each phase to explore

Phases
PHASE 01 Math & Programming Foundations 2–3 months ▾

Before touching AI, you need the core building blocks: Python programming and essential math. This phase is non-negotiable — everything else builds on it.

Math Skills
Linear Algebra Matrices & Vectors Calculus (Derivatives) Chain Rule Probability & Stats Distributions Bayes Theorem
Programming Skills
Python Basics NumPy Pandas Matplotlib Data Structures OOP Jupyter Notebooks
Resources
  • Python for Everybody — freeCodeCamp / Coursera (free)
  • 3Blue1Brown: Essence of Linear Algebra — YouTube (free)
  • Khan Academy Statistics — Khan Academy (free)
  • NumPy & Pandas Crash Course — YouTube / Kaggle (free)
PHASE 02 Machine Learning Fundamentals 2–3 months ▾

The classical foundation of AI. You'll learn how machines learn from data using algorithms, without neural networks yet. Scikit-learn is your main tool here.

Core Concepts
Supervised Learning Unsupervised Learning Regression Classification Clustering Overfitting / Underfitting Cross Validation Feature Engineering
Key Algorithms
Linear Regression Logistic Regression Decision Trees Random Forest SVM K-Means KNN Naive Bayes
Tools
Scikit-learn Seaborn Kaggle Datasets
Resources
  • Andrew Ng: ML Specialization — Coursera (audit free)
  • Hands-On ML with Scikit-Learn — Book by Aurélien Géron
  • Kaggle ML Micro-Courses — kaggle.com (free)
PHASE 03 Deep Learning & Neural Networks 3–4 months ▾

The engine behind modern AI. You'll build neural networks, train them with backpropagation, and work with frameworks like PyTorch and TensorFlow.

Core Concepts
Neurons & Layers Activation Functions Backpropagation Gradient Descent Loss Functions Batch Normalization Dropout / Regularization Optimizers (Adam, SGD)
Architectures
Feedforward NN CNN (Images) RNN / LSTM Autoencoders GANs (Basics)
Frameworks
PyTorch TensorFlow / Keras Google Colab CUDA / GPU basics
Resources
  • Deep Learning Specialization — Andrew Ng, Coursera
  • fast.ai Practical Deep Learning — fast.ai (free)
  • PyTorch Official Tutorials — pytorch.org (free)
  • Neural Networks: Zero to Hero — Andrej Karpathy, YouTube
PHASE 04 Specializations (Pick Your Path) 3–6 months ▾

AI is a big field. Choose a domain that excites you and go deep. You can always explore others later, but depth beats breadth at this stage.

🤖 Natural Language Processing (NLP)
Tokenization Word Embeddings Transformers BERT / GPT Hugging Face Text Classification Sentiment Analysis Named Entity Recognition
👁️ Computer Vision
Image Classification Object Detection YOLO ResNet / VGG Image Segmentation OpenCV Transfer Learning
🎮 Reinforcement Learning
MDPs Q-Learning Policy Gradient OpenAI Gym PPO / DQN
Resources
  • NLP with Transformers — Book by Lewis Tunstall (O'Reilly)
  • CS231n: Convolutional Neural Networks — Stanford (free)
  • Spinning Up in RL — OpenAI (free)
  • Hugging Face Course — huggingface.co (free)
PHASE 05 Large Language Models & GenAI 2–3 months ▾

The frontier of AI today. Understand how LLMs like GPT-4, Claude, and Gemini work, and how to build applications on top of them.

Core Concepts
Transformer Architecture Attention Mechanism Self-Attention Pre-training vs Fine-tuning RLHF Prompt Engineering RAG Vector Databases
Tools & Frameworks
LangChain LlamaIndex OpenAI API Anthropic API LoRA / QLoRA (Fine-tuning) Pinecone / ChromaDB Ollama (Local LLMs)
Resources
  • Attention Is All You Need — Original transformer paper (free)
  • LLM Bootcamp — Full Stack Deep Learning (free, YouTube)
  • Andrej Karpathy: Let's build GPT — YouTube (free)
  • LangChain Documentation — langchain.com (free)
PHASE 06 MLOps, Deployment & Production 2–3 months ▾

Building a model is half the job. This phase teaches you to deploy, monitor, and scale AI systems in the real world — critical for industry roles.

MLOps & Deployment
Docker FastAPI REST APIs Model Serving TorchServe / TFServing AWS / GCP / Azure CI/CD Pipelines
Monitoring & Tooling
MLflow Weights & Biases Data Versioning (DVC) Model Monitoring A/B Testing Streamlit / Gradio
Advanced Topics
Model Quantization Distillation Edge Deployment Responsible AI / Ethics AI Safety Basics
Resources
  • Made With ML: MLOps Course — madewithml.com (free)
  • Deploying ML Models — Full Stack Deep Learning
  • Docker & FastAPI for ML — YouTube tutorials (free)
💡 Pro tip: Build a project at the end of each phase. A portfolio of 6 real projects beats any certificate.