Home / How it works
SchematicPlain version

How RubberAI works.

In short: your recipes and process data go in, the engine learns your chemistry, and you get predicted properties and candidate recipes back — before anything is mixed. Here's the path from data to decision.

/ the pipeline

From your data to a decision.

/01

Bring your data

Upload recipes & process logs as CSV/XLSX — with your own column names. Polymers, fillers, cure system, mixing & vulcanization settings.

/02
yourscanon

Schema mapping

The platform translates your columns into its canonical fields automatically, so you don't have to reformat anything.

/03

The engine learns

Models train on your chemistry and combine with a shared master model — so predictions are useful even before you have thousands of rows.

/04

Results you can act on

Get the predicted cure curve, hardness, compression set and more — each with a confidence band and the key drivers behind it.

/ two ways to use it

Predict forward, or design backward.

Forward prediction

Recipe → properties

You have a compound. The engine tells you how it will behave — before you weigh a single gram.

input
recipe + process
engine
predict
output
Shore‑A 71 · t90 4.8m · CS 14%
Inverse design

Target → recipes

You have a spec to hit. The engine proposes candidate recipes that should land inside it.

target
hardness 70±3 · CS <15%
engine
search
candidates
47 recipes ranked
One dial: master model ↔ your own databalanced for confidence
shared master modelyour custom model
/ the important part

Your recipes stay yours.

Your raw recipes never leave your workspace and are never pooled into a shared model. Any cross‑lab learning happens through federated learningopt‑in and revocable at any time. Read more on security →

/ ready when you are

See it on your compounds.

Request a walkthrough and we'll run the platform against the elastomers you actually work with.

Request a demo