AI Learning Roadmap
Go from your first Python script to building real AI systems. Three phases, no guesswork.
Beginner
No AI experience needed. You'll learn the basic math and coding skills used in every area of AI. Estimated time: 4–6 weeks.
Python Basics
Variables, loops, functions, and classes. Work with data using NumPy and Pandas.
Math for AI
Learn the key math: linear algebra, basic calculus, probability, and statistics.
Exploring Data
Look at data, make charts, handle missing values, and pick useful features.
Classic ML Models
Linear regression, decision trees, k-NN, and k-means using scikit-learn on real data.
Checking Your Model
Split data for testing, use cross-validation, and learn accuracy scores like precision and recall.
Your First Full Project
Build a complete pipeline: load data → clean it → train a model → check results → submit.
Intermediate
Go deeper into neural networks, deep learning tools, and pick a focus area: Computer Vision, NLP, or Reinforcement Learning. Estimated time: 8–12 weeks.
Neural Networks In Depth
How backpropagation works, choosing activation functions, weight setup, and batch normalization.
PyTorch Basics
Tensors, auto-gradients, training loops, GPU speed-up, and loading data.
Computer Vision Track
CNNs, ResNets, reusing trained models, object detection with YOLO, and image segmentation.
NLP Track
Tokenization, word vectors, RNNs, LSTMs, attention, and fine-tuning BERT.
Reinforcement Learning
MDPs, Q-learning, policy gradients, PPO, and DQN with Gym environments.
MLOps Basics
Track experiments with MLflow, save model versions, and build repeatable pipelines.
Advanced
Learn the latest AI models, recreate research papers, and build real-world AI systems. This is where you go from learner to expert. Estimated time: 12–20 weeks.
Transformers & Attention
Build the "Attention is All You Need" model from scratch. Learn about GPT, T5, and CLIP.
Generative AI
VAEs, GANs, and Diffusion Models. Train your own image generator and check the results.
Large Language Models
Fine-tune LLMs, use LoRA, RLHF, prompt tricks, RAG pipelines, and LangChain.
Making Models Smaller
Quantization (INT8/FP16), pruning, knowledge distillation, and deploying with TensorRT.
Recreate a Research Paper
Pick a recent AI paper, build it end-to-end, and present what you learned to the club.
Final AI Project
Design and launch a real AI app: build the API, add monitoring, and set up auto-deploy.