Decision support, not autonomous approval.
We construct and analyze facts to guide lending decisions. The CreditGraph system complies with the Digital Personal Data Protection Act (DPDP) and RBI guidelines, putting control back into human hands.
Review Framework100% Auditable Logs
Every credit decision flag outputs the exact speech recording, receipt ID, and bank confirmation that generated it.
Our 4 Compliance Pillars
Built from the ground up for strict RBI compliance and DPDP requirements.
01. Explicit Consent
Merchants opt-in directly via signed WhatsApp templates. Consent is clear, revocable, and audited. No hidden background scraping.
02. Data Security
AES-256 encryption at rest and in transit. All user data is hosted within Indian sovereign servers, strictly adhering to RBI localization rules.
03. Human-in-the-Loop
We generate explainable underwriting parameters, not direct credit approvals. An institutional human underwriter always makes the final call.
04. Linguistic Fairness
Linguistic NLU models are audited weekly for dialect bias to ensure shop owners using regional accents (Tamil/Hinglish) get fair credit indexing.
Immutable Decision Log Audits
To ensure 100% compliance transparency, CreditGraph generates immutable decision trails. Lenders can click any past loan recommendation and see the raw speech, OCR, and banking data variables that drove it.
This prevents credit decisions from turning into unexplained AI black boxes, protecting lenders during regulatory audits.
> Fetching profile: Sharma Electricals (ID_902919)
> Consent Verified: WhatsApp opt-in signed (Token: CNS_90192)
> Analyzing audio ledger log CNS_REC_0815.mp3:
- Transcript: "Ramesh ne 35 hazar ka maal liya tha..."
- Entities: Ramesh Kumar (Buyer), ₹35,000 (Invoice)
> Cross-matching UPI statement reference:
- Matched transaction ID: 6231920194 (₹10,000 received)
- Reconciled Receivable updated to: ₹25,000
> Underwriting scoring calculation triggered:
- Rec stability: 82% | Concentration: Moderate (31%)
- Recommended Limit: ₹4.0L - ₹6.0L
> State check complete. Audit record sealed.
Consent & Encryption Architecture
How user data flows from instant opt-in validation to long-term audit seals.
Continuous Bias Assessment Logs
To comply with fair lending guidelines, our linguistic NLU is audited weekly. We measure parsing error differences across regions to ensure dialect usage doesn't lead to higher default estimations.
The logs show consistent validation parity across business dialects, ensuring equitable capital access.
| Dialect | Audit Date | Word Error Parity | Status |
|---|---|---|---|
| Hinglish (North) | 12 Aug 2026 | 1.12% Deviation | Pass |
| Tamil (South) | 12 Aug 2026 | 0.94% Deviation | Pass |
| Telugu (Agri) | 11 Aug 2026 | 1.34% Deviation | Pass |