What jAIn predicts
One column of one table, and the three tasks that can be: binary classification (two outcomes), multiclass classification (several), and regression (a number). jAIn infers which of the three your target column is.
What it does not do — anything temporal, text, images, audio, video, or a shape that is not one table with one target — is listed in full on the product page. It is worth two minutes before you upload anything.
What a session looks like
- Sign in with a magic link or a password.
- Upload a CSV or TSV and say what you want predicted.
- The run is queued. One session runs at a time per account.
- A container starts on an isolated network and does the work.
- Results are captured before the container stops, so they stay readable afterwards.
What you get back
A trained model, a report describing how it was found and how well it does, and predictions for the rows you asked about. All three are downloadable for as long as they are kept.
How long results are kept
The file you upload is deleted automatically after 30 days. What the run made from it — your report, your trained model and its evidence receipt — is kept. You can delete a session, or your whole account and all of its data, at any point. The details are on the privacy page.
This number is not written by hand — it comes from the same configured value the app enforces, so the promise on this page cannot drift away from the deletion that actually runs.
Running it yourself
You can install the same workspace and run it on your own machine, but be clear about what that moves: the interface, not the compute. The search still runs on our hardware and your table is still uploaded, so it is not an offline mode and there is not one today. Running locally covers the install, the product page covers where a private deployment has got to, and the security review covers what isolates a hosted run.