Useful resources

A short, curated selection to progress in AI: beyond the course. The course remains the core; these links complement the theory, the documentation and the practice.

The olympiad (IOAI / OFIA)

The target competition and its official syllabus.

IOAI: official siteInternational Olympiad in Artificial Intelligence: format, editions, teams.IOAI SyllabusThe official program of expected knowledge. It is our reference baseline.OFIA (France)The French Artificial Intelligence Olympiad, the national selection for the IOAI.

Communities and practice contests

To train with other AI olympiad candidates, share and compete regularly.

SOTAA community to prepare for the IOAI: independent, free and student-run. A Discord with several hundred members, a guide that follows the IOAI syllabus, a checklist of hundreds of past olympiad tasks, contests and mentoring by former contestants.AICCMonthly IOAI-style practice contests, run on Kaggle by a community of former contestants, with a Discord of several hundred members: vision, NLP, audio, classical ML. Past tasks, with our editorials, are also on the Problems page.

Before you start: programming

If you can't code yet, learn the basics of Python first, then come back to Level 1.

France-IOI: programming courseThe best French-language resource to learn Python and algorithmics from scratch, through practice.

Tool documentation

The libraries you will use throughout.

PyTorch: tutorialsThe platform's deep learning framework. Official tutorials from the tensor to the trained model.NumPyVectorized numerical computing. Documentation and getting-started guide.pandasTabular data manipulation (DataFrames).scikit-learnClassic ML: regression, classification, metrics, validation.Hugging Face: LearnPractical courses on transformers, NLP and diffusion.

Reference online courses

Excellent, free courses to build intuition and rigor.

3Blue1Brown: Neural NetworksThe best visual intuition for neural networks (and for linear algebra / calculus).Karpathy: Neural Networks: Zero to HeroBuild a network, then a GPT, by hand, line by line. A must.fast.ai: Practical Deep LearningA top-down, very practical approach to train real models quickly.Dive into Deep Learning (d2l.ai)Free interactive book, theory + code, from the MLP to transformers.Stanford CS231n: VisionThe reference course in computer vision (CNN, detection, segmentation).Stanford CS224n: NLPThe reference course in modern NLP (embeddings, attention, transformers).

Practice for real

Get your hands dirty on real data and real competitions.

Kaggle: LearnPractical micro-courses (Python, ML, DL, pandas) with built-in notebooks.Kaggle: CompetitionsData science competitions to practice on real problems.

Is an excellent resource missing? Let us know via the feedback button: this list is intentionally short and hand-picked.