49 scikit-learn models, full cross-validation, instant Plotly visualisations. No signup. No data upload. Your CSV never leaves your machine.
Free · runs on Pyodide · first load ~10 s, then cached
All the power of scikit-learn, none of the ops. The whole runtime ships in the page.
Python + scikit-learn compiled to WebAssembly via Pyodide. No server, no Docker, no Render bill.
CSV uploads never touch a network. Everything runs inside your tab — air-gapped by default.
27 regressors and 22 classifiers across linear, tree, ensemble, SVM, neighbours and neural nets — compared on the same chart.
3 / 5 / 10-fold CV with R², MSE, RMSE, MAE for regression; accuracy, F1, precision, recall, ROC for classification.
Download any trained model as a real .joblib file ready to load() in your own Python project.
Iris, Wine, Breast Cancer, Digits, Diabetes, Airfoil Self-Noise — click and train, no upload needed.
No notebooks. No environment setup. No pip install.
Drop a CSV or pick a sample dataset. Classification or regression is auto-detected.
Choose your target column and predictors from a clean side panel.
Select one or many. Cross-validation runs automatically.
Inspect residuals, ROC, confusion matrices — then download the winning model.
Two task types, six families each. All run client-side.
One learner.py runs behind all of them — as a page, as a Jupyter side panel,
as a VS Code sidebar. Pick the one you already have open.
A tab and nothing else. Pyodide loads scikit-learn into the page and your CSV never leaves the machine.
Open the app →
A side panel next to your notebooks: dataset, models, runs, artifacts. Models are fitted
in the kernel — Pyodide in the browser, or your own Python in JupyterLab — and the
generated pipeline.py opens as a real file you can run.
The same four sidebar sections and a plots editor, training against a local Python environment the extension can set up for you.
Get the extension →
Open a tab, train a model, ship the joblib — or bring the same panel into the notebook you already have.