Learn data science and AI engineering. One concept at a time.

The fastest way to go from zero to writing real SQL. Query, filter, aggregate, and join across the Bean & Brew coffee-shop dataset, with animated explainers that show exactly how each query runs and a live editor that checks every answer instantly. No installs, no prerequisites.

Level up to the SQL that analytics roles actually demand. Master CASE logic, subqueries and CTEs, and window functions through interactive animations that reveal what happens under the hood, then prove it on real data in a live, in-browser editor. Built to make hard queries feel obvious.

Master the single highest-leverage advanced SQL skill. OVER(), ranking, LAG/LEAD, frame clauses, the multi-dialect quirks that trip everyone up, and the performance patterns that make analytical queries fast, across eight animated explainers and 50+ hands-on challenges in a live editor.

From hypothesis to production-grade experiment. Master the statistics, engineering, and judgment behind experiments that actually hold up, power and sample size, randomization and SRM, plus modern methods like CUPED, sequential testing, and bandits, all in browser-native scipy & statsmodels.

How large language models really work, from BPE tokens through RoPE attention, scaling laws, DPO/GRPO, KV caches, and mechanistic interpretability. Eight modules and eight interactive animations that take you from curious beginner to confident practitioner, no math degree required.

Build production retrieval-augmented generation from first principles, BM25, dense embeddings, hybrid search, chunking, reranking, ANN indexes (IVF/HNSW), HyDE, and rigorous evaluation. Every technique measured before-and-after on one realistic corpus. Runnable Python, no API keys.

Build AI agents from scratch, the agent loop, tool use, planning, reflection, memory, multi-agent systems, evaluation, and guardrails. The model is a deterministic stand-in, so every piece of agent machinery runs for real on the Helpwell support agent. Runnable Python, no API keys.

Evaluate LLM, RAG, and agent systems for real, error analysis, deterministic & semantic scorers, LLM-as-judge alignment (Cohen’s kappa), confidence intervals, McNemar regression tests, and a CI gate that blocks bad deploys. Every scorer, statistic, and gate runs live on the Helpwell assistant.

Build a working language model from scratch in your browser: a tokenizer, embeddings, attention, the transformer block, and the sampler, all in pure runnable NumPy. Then watch StoryByte, a real 1M-parameter GPT we trained on TinyStories, write an original story one token at a time. No PyTorch, no GPU.

The most visual way to learn Python for data science. Master variables, control flow, functions, data structures, and file I/O through interactive animations and a live editor that runs real Python in your browser, with instant feedback on every exercise. No installs, no prerequisites.

Go from spreadsheets to professional data wrangling. Master DataFrames, selection, groupby, merging, reshaping, and time series, plus the performance patterns that separate hobbyists from pros, on the Acme SaaS dataset, with animated explainers and real pandas running live in your browser.

Build real statistical intuition, not just memorized formulas. Master mean, median, and mode, spread and variability, distributions, and clear data visualization, through interactive visualizations that make every concept finally click. The foundation everything else in data science stands on.

The language of uncertainty, from coin flips to the math behind every ML model. Conditional probability, random variables, the distribution zoo, Bayes’ theorem, and the limit theorems (LLN & CLT), across 8 modules with interactive animations, runnable Python, and the Streamora running dataset.

The supervised-learning loop the way working data scientists run it, regression, classification, regularization, cross-validation, and leak-free pipelines. Every model trains live in your browser on the Lendly P2P-lending dataset, built on scikit-learn 1.3. Theory you can actually run.