Explainable AI · Digital Lending

Every lending decision, entered in the open.

A passbook is a ledger you're allowed to read. This is the same idea applied to a credit model — a demo eligibility engine that shows its SHAP contributions and plain-language reasoning for every decision, not just the verdict.

Check your eligibility → Read the methodology
CreditAI · Model Passbook
XAI
CERT
Digital Lending
Credit Decision Register
Prepared for  Explainable underwriting, demonstrated live
Entries 01–04

How the model reads a file

Four inputs are entered, weighed, and posted — the same order a credit officer would work through, made visible at every step.

01

Applicant profile

Identity, employment type and tenure — the baseline signal for repayment stability before any loan terms enter the picture.

Stability
02

Loan terms requested

Amount, purpose and tenure. The model checks these against income before it ever looks at credit history.

Exposure
03

Financial standing

CIBIL score, existing EMIs and prior defaults are combined into a debt-to-income figure — the single strongest predictor in most digital lending models.

Risk
04

Decision, posted with reasons

A verdict is issued alongside a SHAP breakdown and a plain-language (LIME-style) explanation — debited and credited factors, laid out like a ledger balance.

Disclosure
Note on method

Why the reasons matter as much as the result

Most credit-scoring models are correct more often than they are understood. Two techniques close that gap, and this demo uses a simplified version of both.

Global reasoning

SHAP contributions

Each factor — credit score, DTI, income, tenure — is assigned a signed value showing exactly how much it pushed the decision toward approval or rejection, and by how much relative to the others.

Local reasoning

Plain-language rationale

The largest contributors are translated into sentences a loan officer — or an applicant — can actually act on, rather than a table of coefficients.

Live entry

Open the ledger

Enter a sample profile below. The scoring logic runs entirely in your browser — no data leaves this page.

New Application Entry
Model active
1 · Personal
2 · Loan
3 · Financial
4 · Decision
Posting to ledger, weighing contributions
Default probability
Risk category
Estimated EMI
SHAP ledger — feature contributions Dr = adds riskCr = reduces risk
Plain-language rationale
About this demo This is an academic/portfolio project built to illustrate explainable AI concepts (SHAP-style attribution, LIME-style plain-language reasoning) in a digital lending context. The scoring logic is a deterministic, rule-based simulation for demonstration only — it is not a trained model and does not connect to any credit bureau or lender. Do not use this to make real financial decisions.