How do you explain a kind of lending that has never existed here before?
Leaf underwrites businesses on proprietary ecosystem data instead of credit scores. Our customers are MSMEs — often with no credit history at all. A bank looks at them and sees nothing to lend against; we look at their ecosystem and see everything: where they work, what they deliver, and how reliably they get paid. Our first ecosystem was GeM, the Government e-Marketplace, where we underwrite sellers on their actual trading history. The site speaks to two audiences at once: businesses seeking credit, and the ecosystems that host thousands of them. For an ecosystem, Leaf is plug-and-play lending infrastructure — live in about a week, with Leaf running the entire stack, integrating with their ERP, and sharing the upside.
A loan in under a minute — designed and shipped at the speed of AI agents.
India's businesses live on the web, not mobile apps, so the journey is laptop-first. Onboarding is deliberately light: the customer shares KYC and business details, and with their consent we pull everything else from India's centralized data infrastructure — then underwrite and disburse. The build matched the product's speed. I drew skinny wireframes in Figma, used Claude Code agents to generate the high-fidelity design system, and handed finalized designs straight to AI agents on engineering — built and hosted on Vercel. Live in one week, with customers using it, then continuous beta iteration to design out the drop-offs.
The Internal Platform
The most intense build of the three: regulation, operations, and risk in one designed experience.
As an RBI-regulated lender, every operation carries deep audit trails. Working directly with the founders, I framed the platform through four lenses — business, operations, compliance, and company health — and designed the full suite: a loan origination funnel with fully attributable drop-offs, a dashboard of the book's vital statistics, loan management where every repayment has its own trail, a built-in CRM that turns compliance records into a working workflow, P&L and unit-cost analysis, bureau reporting and reconciliation, and the newest lens — risk, where regressions and time-based analysis feed our proprietary AI model to read the live health of the portfolio.
Final Thoughts
$500K disbursed and 800+ businesses borrowing — in the first six weeks.
Minimal friction, and one week from design handoff to a live product. Three surfaces, one system: a marketing site that explains a category that didn't exist in India, a customer journey that goes from landing to loan in under a minute, and an internal platform where regulation, operations, and risk live in one designed experience. Built design-first, and shipped at the speed of AI agents.
Next projects.
(2016-26©)
