Digital Lending Platform

Loans took days, and borrowers dropped out along the way. The NBFC lends to small businesses and salaried borrowers across eastern India, through branches and loan agents.

Paper files were couriered to head office. KYC documents and bank statements were checked line by line by hand, and underwriters spent more time typing figures than judging risk. The new journey also had to be digital lending done right: built around the RBI’s digital lending rules from the first screen, not patched afterwards.

We mapped every step of the loan journey against the RBI’s digital lending rules with the NBFC’s compliance team before designing anything. They shaped the product as much as any user need:

• a Key Fact Statement with the annual percentage rate and total cost, shown before the borrower signs;
• disbursal straight to the borrower’s own bank account, and repayments straight to the lender;
• a cooling-off period in which the borrower can exit the loan without a penalty;
• need-based data, collected with consent, with no access to the phone’s contacts, call logs or photos;
• data stored in India, and every consent and decision kept on record;
• the grievance officer’s details visible in the app and on the website.

Borrowers apply in the app or on the web. They verify their mobile number, enter their PAN, and complete e-KYC through Aadhaar OTP or DigiLocker. Their CKYC record is checked and updated, and their bank account is verified with a small test transfer before any money moves.

Where the loan needs it, video KYC follows in the same sitting: a live call with a trained agent in India, with liveness and face-match checks, a location check, and a random question to confirm the borrower is present. Most borrowers finish KYC without leaving their chair.

Bank statements arrive through the Account Aggregator framework, with the borrower’s consent, straight from their bank. Where that isn’t possible, borrowers upload them.

AI then reads statements, salary slips, income tax returns and GST returns. It extracts the figures underwriters need, such as income, obligations, bounced payments and turnover, and flags anything that doesn’t match, like a salary credit that differs from the slip. People still decide. AI only removes the data entry and points to what deserves a second look.

The credit bureau report is pulled automatically. A rules engine that the credit team configures itself runs the scorecards, eligibility and policy checks, and works out the loan amount and terms each applicant qualifies for.

Applications that need judgement go to an underwriter with the whole case on one screen: KYC, documents, extracted figures, bureau report, rule results and flags. Deviations from policy follow an approval matrix, and every decision records its reasons. When the credit team changes a policy, it goes live without waiting for a software release.

Approved borrowers see their offer and the Key Fact Statement, sign the agreement with Aadhaar e-sign, and set up repayments by e-NACH. The loan is disbursed straight to the borrower’s own bank account, and the borrower receives the signed agreement and repayment schedule in the app.

Loan servicing runs on the same platform. Borrowers see their repayment schedule, pay early or part-prepay, download statements and request a closure letter. Reminders go out before each instalment, and the collections team sees missed payments the day they happen. Loan data is reported to the credit bureaus every month.

All borrower data is stored in India, on AWS’s Mumbai region, and encrypted in transit and at rest. Each staff role sees only what it needs, and every step, consent and decision is logged for audits. Before launch, the platform passed independent vulnerability assessment and penetration testing, and every finding was fixed.

The core platform is Java and Spring Boot on PostgreSQL, with Apache Kafka carrying events between services, from KYC to credit to servicing. The borrower app is built in Flutter and the staff workbench in React. Document reading uses Amazon Textract and Amazon Bedrock, within AWS in India, and Python services run the analysis of bank statements. The platform launched one loan product at a time, starting with salaried personal loans, then loans for small businesses.

Borrowers now apply from their phone, complete KYC in one sitting, and receive money straight in their bank account. Underwriters spend their time on judgement instead of typing, the credit team changes policy without a release, and every consent and decision is ready for an audit.