zypl.Foundationdemoyour own file or a public example
Close the Outcome Gap
Credit decisions from day zero, measurably better with every matured loan, on your own servers.
Day-Zero · before any outcome
A reviewable scorecard from your documentation and feature statistics: points, grades and reason codes before the first repayment.
zypl.Foundation · with matured outcomes
A labelled book goes in; ranked scores and ROC and PR curves come out. The model learns the book in context — no training run per product.
+2.1
Gini points at 100 matured loans
0
borrower records in pretraining
~50 ms
per application on a 16-thread CPU
Research results, not results of this demo: zypl labs paper under review and serving report v3.1. +2.1 is against TabICLv2 with eight members, ranking quality not calibration; ~50 ms is after a one-time portfolio encoding, at a 12,000-row context.