Anonymized compositeFintech Pune 10 weeks to production

A 22-Person Fintech Startup in Pune Cut KYC Processing from 3.5 Hours to 11 Minutes

Document AI pipeline that extracts, cross-verifies, and scores KYC submissions against RBI-compliant rules in 11 minutes — versus the 3.5 hours a two-person verification desk was taking per application.

KYC processing time

3h 30m → 11 min

95% faster

Industry

Lending / NBFC

Location

Pune, Maharashtra

Timeline

10 weeks to production

Client: A 22-person fintech startup in Pune building a lending platform for Tier-2 cities. Client name withheld. This page is an anonymized composite of recurring implementation patterns. Figures are representative internal examples, not independently audited results or guarantees.

Key Facts

Industry: Fintech — NBFC / lending
Location: Pune, Maharashtra, India
Problem: KYC processing took 3.5 hours per applicant; manual document review at scale
Solution: Custom OCR + LLM document pipeline with risk scoring
Timeline: 6 weeks to pilot, 12 weeks to full rollout
Headline outcome: KYC time cut from 3.5 hours to 11 minutes (~95% reduction)
Compliance: DPDP Act-aligned data minimisation and consent capture

The Challenge

▸Three compounding tools — Claude Code for business logic, GitHub Copilot for general coding, Cursor for review — each charging per-user, none aware of the startup's credit-scoring logic or RBI compliance layer.
▸Two senior engineers had effectively become full-time reviewers of AI-generated code rather than builders.
▸KYC verification desk was the operational bottleneck — applications came in faster than the team could verify them, pushing approvals from 24h to 72h during promotion periods.
▸False-positive flag rate on the rules engine was 18%, meaning nearly one in five genuine applicants got held for manual review.

What We Built

1Built a document intelligence layer that ingests Aadhaar, PAN, bank statements, and income proof in any combination of formats (image, PDF, scanned).
2Cross-verification against public Digital India APIs (where available) plus a confidence-scored internal risk model trained on the startup's own historical approval outcomes.
3Compliance-aware flagging: any submission with RBI red-flags routes immediately to a named human reviewer with the specific flag surfaced — no silent auto-denial.
4Audit trail: every decision, model version, and input document is retained with cryptographic integrity for 7-year regulatory retention.
5Dashboard for the compliance officer showing model drift, override patterns, and queue health in real time.

Representative Outcomes

These internal examples illustrate how success can be measured. Actual outcomes depend on baseline performance, data quality, integration scope, operating controls, and user adoption.

3h 30m → 11 min

Average KYC processing time

95% faster end-to-end

18% → 4%

False-positive flag rate

Fewer genuine applicants held

72h → 6h

Application-to-decision time

During promotional surge periods

3 → 1

Dedicated verification headcount

Two engineers redeployed to lending-product work

Technologies Deployed

Document AI with layout-aware extraction
Custom LLM for document reasoning (fine-tuned on Indian KYC edge cases)
DigiLocker + Digital India API integrations
On-prem deployment for data sovereignty

Services Used

Nlp Custom Gpt Solutions Ai Powered Erp Tools

Want a similar engagement on your operations?

Book a free 30-minute AI workflow review. We'll identify the processes where AI genuinely moves numbers for your business and hand you a realistic delivery plan.

Start a Conversation

More Case Studies

Healthcare