HACKATHON · FINTECH AI

VerdeScore

Scores green loans for EU Taxonomy and LMA compliance in milliseconds instead of days.

↓ SCROLL

The problem

Green loan review is manual: an analyst reads the application, checks it against EU Taxonomy thresholds and LMA Green Loan Principles by hand, and the whole cycle takes 10-12 days. Built for the LMA EDGE Hackathon 2026, where the brief was to make that review fast without making it careless.

The approach

VerdeScore reads the loan application, extracts and embeds the claims with a semantic model, then validates each one against EU Taxonomy thresholds and LMA Green Loan Principles. Every score comes with an explainable audit trail, so a human reviewer sees why the system decided what it decided rather than trusting a black box.

Impact

13ms
Per-loan scoring time
10-12 days
Prior manual cycle
66,461x
Claimed speedup

What it does

Semantic claim extraction

Reads the loan document and pulls out environmental-impact claims automatically.

Taxonomy + LMA validation

Checks each claim against EU Taxonomy thresholds and the LMA Green Loan Principles.

Explainable scoring

Every score ships with an audit trail a reviewer can actually read.

Stack

React 18ViteFramer MotionNode.jsExpressSQLiteDeepSeek API
LMA EDGE 2026Hackathon
13msPer loan
DeepSeek AICore model
SoloRole