Resources & references

Curated, low-noise links. Skip the AI-influencer YouTube feeds; these are the materials that actually move the needle.

Official

usaaio.org — Contest home, registration, syllabus, past problem archive, important dates.
IOAI official site — International Olympiad in AI; problems and country results.

3Blue1Brown — Essence of Linear Algebra (YouTube). The geometric intuition for vectors, matrices, determinants, eigenvectors. Watch it once a year.
3Blue1Brown — Essence of Calculus. Same treatment for derivatives, chain rule, integrals.
Mathematics for Machine Learning — Deisenroth, Faisal, Ong. Free PDF at mml-book.github.io. The exact subset of math used in ML, no more.
Khan Academy — fill in any high-school gaps in algebra, probability, calculus.

Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow — Aurélien Géron. Read Parts I & II (skip the TensorFlow chapters; do PyTorch on the side).
An Introduction to Statistical Learning (ISL) — James, Witten, Hastie, Tibshirani. Free PDF at statlearning.com. The conceptual foundation for every model in scikit-learn.
scikit-learn user guide at scikit-learn.org. Worth reading the supervised-learning chapter cover-to-cover.
Kaggle Learn — short interactive lessons on pandas, ML, feature engineering. Free.

Deep Learning — Goodfellow, Bengio, Courville. Free at deeplearningbook.org. Dense; read Chapters 6–9 first.
PyTorch official tutorials at pytorch.org/tutorials. Start with "Learn the Basics" and "Quickstart."
Karpathy's "Neural Networks: Zero to Hero" series (YouTube). Build everything from scratch in pure Python, then PyTorch. Watch micrograd and makemore first.
fast.ai Practical Deep Learning for Coders at course.fast.ai. Top-down, hands-on. Free.

"Attention Is All You Need" (Vaswani et al., 2017). The original transformer paper. Read it after watching Karpathy's GPT-from-scratch video.
Karpathy "Let's build GPT: from scratch, in code, spelled out" (YouTube). Best single resource for understanding transformer training end-to-end.
The Illustrated Transformer — Jay Alammar at jalammar.github.io. Visual companion to the original paper.
Hugging Face course at huggingface.co/learn. Practical fine-tuning with the transformers library.

Kaggle — Getting Started competitions (Titanic, House Prices, MNIST) for fundamentals, then Featured competitions when you're ready.
AIcrowd — Hosts research-style ML competitions.
Codeforces / AtCoder — Not AI, but algorithm contests sharpen the coding speed you'll need under contest pressure.
Hugging Face Spaces — Ship a working demo of every model you train. Cheap motivation.

Python 3.11+, venv or conda for environments.
JupyterLab for exploratory work, VS Code for serious editing.
Git + GitHub for the portfolio repo. Commit every notebook.
Weights & Biases (free for students) for experiment tracking once you start running many DL experiments.
Google Colab as a free GPU backup. Useful when local hardware can't keep up.

Term	What it means
Epoch	One full pass through the training data
Batch / mini-batch	Subset of training data used in one gradient step
Learning rate (LR)	Step size for the optimizer update
Logits	Raw, unnormalized model outputs before softmax / sigmoid
Cross-entropy	Loss for classification: −Σ y log(p)
Backpropagation	Chain-rule algorithm for computing gradients through a network
SGD / Adam / AdamW	Optimizers (update rules for parameters from gradients)
Overfitting	Training loss low, validation loss high — model memorized noise
Regularization	Techniques to reduce overfitting (dropout, weight decay, etc.)
Embedding	Dense vector representation of a discrete token / category
Attention	Mechanism that weighs input positions when producing each output
Self-attention	Attention where Q, K, V all come from the same sequence
Pre-train / fine-tune	Train on big general data, then adapt on small task-specific data
LoRA / PEFT	Parameter-efficient fine-tuning — train a small adapter, freeze the base
Inference	Running a trained model on new inputs (no gradient updates)
Token	Discrete unit fed to a language model (subword, character, or word)