M· Small teamSaaS Tools

Model Lab

Train regression and classification models in your browser and score them honestly, from linear and logistic to trees, forests and boosting. Private, no account.

K3N8H Verified

✓ Real preview shipped
✓ License terms specified
✓ Setup/usage instructions included
✓ Actually run before listing

Disclaimer: Model scores are estimates of how well a model would do on similar data, not guarantees. A model can be over-fit or leaky, so read the warnings before trusting a result. This is not financial, medical or legal advice.

Turn a table into a prediction. Pick the column you want to forecast or classify, and the tool works out whether it is a number, a yes/no or several classes and why, suggests which other columns to use as features, and warns you about small samples, unbalanced classes, and a feature that is almost the answer (leakage). Choose one of ten methods - a no-learning baseline, linear, ridge, lasso, logistic, nearest neighbours, a decision tree, a random forest, gradient boosting, or Naive Bayes - set its few controls, and train. The preparation is fitted on the training rows only, so it applies unchanged to new data. You then read the held-out scores: mean error and R-squared for numbers, accuracy, AUC and calibration for classes; for a number the error it is likely to be off by, for a tree the questions it asks, and for the linear models the coefficients with p-values or odds ratios. Save a model to a file and reload it later to apply to a brand-new table, or hand the cleaned table back to the Data Analyst Workbench. Runs entirely in your browser; the data never leaves your device.

Runs where you work

desktopmobileoffline

Built for a private workflow: local-first execution, no forced cloud dependency, and a clear path to use it on a laptop or phone without an API key.

Included value

Ten methods: baseline, linear, ridge, lasso, logistic, nearest neighbours, decision tree, random forest, gradient boosting and Naive BayesAutomatic detection of a number, yes/no or multi-class target, with the reason shownHonest data checks: tiny samples, class imbalance, rows-per-feature and leakagePreparation fitted on the training rows only, so it applies unchanged to new dataHeld-out scores: R-squared and mean error, accuracy, AUC and calibrationThe likely error band for a number, the drawn tree, and coefficients with p-values or odds ratiosCross-validated penalty choice for ridge and lasso, and out-of-bag scores for the forestSave a model to a file, reload it and apply it to a new table; hand the table back to the Data Analyst Workbench

Each product is designed to feel personal, useful, and complete without turning your work into an external dependency.

Free tools that go with this

These run in your browser, need no account, and keep what they remember on your own device.

Explain Like I'm 5Plain-language explanations at five reading levels, all on your device.Code LabDescribe a task or paste code: run it, translate it, explain it, practise it.TerminalA practice terminal with a virtual file system, shell commands, a file editor, and node and python runners for beginner programs.

Preview

Open full preview →

This product is a live, in-browser experience — not a file to preview.

Free to play right now, in this browser, no download or account required.

Play now →

Runs entirely in your browser with no API key or account. Open a table or the practice data, train a model and read the held-out scores; nothing is uploaded.

What's included

Live tool at /play/data-scientist (free, unlimited use, no account)
Commercial license: white-label/embed this tool on your own site

Built with

TypeScriptReact

Runs in your browser: nothing to install

Engine in src/lib/data-scientist, verified by scripts/test-data-scientist.ts (502 checks).

License & pricing

Try it before you buy. This tool is free to use on the site: open it.

Free to Use

Free

For your own work or your own business. No client work, and no resale or redistribution of the files.

Embed / White-Label License

$49

Use across client work or resell the output. No reselling the source files/template itself.

What these licenses mean →

Checkout for the paid licence is not open yet. The free version works now: open it.

Refund terms are set per product, and checkout through Whop or Gumroad has its own buyer protection. Questions before you buy? Ask in the chat or on the feedback page; we'll help where we can.

SaaS Tools — publishing standard

Required deliverables: Source code or hosted access, README, Setup guide, Run proof
README required
Proof-it-runs evidence required
Personal vs. commercial license tiers offered

State real feature scope and limits; no vaporware claims.