Platform · ML Unified
Four trained models. One app.
Fill a form, get a prediction.
ML Unified is a tabular-prediction world inside AIRaML. Pick one of four models, fill a form the app builds automatically from that model's schema, and get a live prediction — with class probabilities or a predicted value — all from one schema-driven FastAPI backend.
You enter — Titanic Survival
- Passenger class
- 3
- Sex
- female
- Age
- 28
- Fare
- £7.90
- Siblings / spouse
- 0
The model predicts
Survived
71% confidence · Gradient Boosting
An illustrative example. The model is trained on the real Titanic dataset (82.5% held-out accuracy) — a demo, not advice.
One app, four models
Pick a model, get a prediction
Three classifiers and a regressor, each trained on a real dataset and served from one schema-driven backend. Each card shows the model's real held-out metric.
Iris Species
Classify an iris into one of three species from four petal and sepal measurements.
Titanic Survival
Predict whether a passenger survives, from class, sex, age, fare and family aboard.
Diabetes Risk
Estimate diabetes risk from eight clinical measurements such as glucose and BMI.
Insurance Premium
Predict an annual insurance premium from age, BMI, smoking status and region.
How it works
Pick → fill → predict
Pick a model
Choose one of the four models from the sidebar. The form on the right rebuilds itself from that model's feature schema — the right fields, labels and ranges.
Fill the form
Every numeric field shows its valid range and step; categorical fields become dropdowns. Sensible defaults are pre-filled, so you can predict right away.
Predict and read
The trained model runs on the FastAPI backend. Classifiers return the class plus per-class probabilities; the regressor returns the predicted value.
Honest scope
Real models, with the caveats stated
These are real trained models on real data — and they are compact models on small public datasets, built to show an end-to-end ML app cleanly, not to give advice.
LiveWhat it does
- ✓Four trained models — Iris, Titanic, Diabetes, Insurance
- ✓Forms generated automatically from each model's schema
- ✓Live predictions from a real FastAPI backend
- ✓Class probabilities for classifiers, a value for the regressor
- ✓Honest held-out metric shown for every model
- ✓One microservice shared with EDA Explorer + Vision
CaveatsWhat to keep in mind
- —Compact models on small public datasets — a demo, not a product
- —Not medical, financial or actuarial advice
- —A prediction is only as good as its inputs and training data
- —Inputs must stay within each field's stated range
- —Probabilities are the model's estimate, not ground truth
Make your first prediction.
Pick a model, fill the form, and see the prediction — no setup, free.
Open the platform