⚡ Pure Rust · WebAssembly · Zero Dependencies · Zero Backend

Tabular ML Demo

Ten real datasets. Ten trained neural networks. All running live in your browser — compiled from hand-written Rust to a single 125 KB WebAssembly binary. No server. No Python. No cloud. Move sliders, get instant predictions.

Classification

🌸

Iris Species

Classify iris flowers from petal and sepal measurements. The original ML benchmark — 3 species.

4 features3 classes 150 rows98.7% acc

🐧

Palmer Penguins

Identify Adelie, Chinstrap, and Gentoo penguins from bill and body measurements.

4 features3 classes 342 rows99.4% acc

🌾

Wheat Seeds

Classify wheat varieties (Kama, Rosa, Canadian) from 7 X-ray geometric kernel measurements.

7 features3 classes 210 rows99.5% acc

🍷

Wine Quality

Rate Portuguese red wine quality (low/medium/high) from 11 physicochemical properties.

11 features3 classes 1,599 rows80.9% acc

🩺

Pima Diabetes

Predict diabetes onset in Pima Indian women from 8 health measurements. Binary classification.

8 features2 classes 768 rows93.0% acc

❤️

Heart Disease

Detect heart disease from 13 clinical variables — age, BP, cholesterol, ECG results, and more.

13 features2 classes 297 rows96.3% acc

🔬

Breast Cancer

Classify tumours as malignant or benign from 30 nuclear features derived from cell images.

30 features2 classes 569 rows99.3% acc

🚢

Titanic Survival

Predict passenger survival from class, sex, age, family size, and fare paid.

6 features2 classes 891 rows86.9% acc

Regression

🚗

Auto MPG

Estimate fuel efficiency (miles per gallon) from engine specs and model year. 1970–1982 cars.

6 featuresregression 392 rowsRMSE 1.95 mpg

🏠

CA Housing Prices

Estimate median house values in California census tracts from location and demographics.

8 featuresregression 20,433 rowsRMSE ~$52k

One 125 KB WASM binary, ten model files (1–10 KB each). The entire ML stack — tensor math, backpropagation, normalizer, binary format — written from scratch in Rust, zero external dependencies. Each model file embeds its own feature metadata so the UI builds itself dynamically.

View source on GitHub