Credit scoring was built for salaried employees.
Gig workers don't exist in that world.
India has over 15 million gig workers — delivery riders, ride-hail drivers, freelancers, platform labourers. They earn real money, they spend consistently, they repay debt. But traditional lending systems cannot see them. No salary slip. No fixed employer. Thin or absent bureau scores. The standard underwriting model rejects them before the first question is asked.
Aamdhanee's thesis: spending behaviour is a stronger proxy for repayment ability than employment history. If you can map a borrower's cashflow consistency, platform earnings patterns, and spending habits with enough resolution, you can underwrite them with confidence — bureau score or not. The thesis is right. The challenge is engineering it to work at scale.
"The bureau score is a proxy for trust. Spending data is trust itself — a direct record of financial discipline over time. ZeroOne built the system that makes that argument in real-time, on every loan application."
Five engineering modules.
One decision in real time.
AI-Powered Underwriting Intelligence
The core engine that replaces bureau score dependency. Ingests spending-habit signals, runs them through behavioural risk models, and produces a composite underwriting score calibrated to gig-worker income patterns rather than salary-based proxies. The model learns continuously from repayment outcomes to sharpen accuracy over time.
Gig Worker Data Pipeline
Unified ingestion layer for platform earnings (Swiggy, Zomato, Ola, Uber, Dunzo, Urban Company), UPI transaction records, wallet activity, and cashflow timelines. Normalises heterogeneous data formats from multiple platforms into a single structured borrower profile. Handles real-time and batch modes.
Loan Decision Engine
Real-time risk scoring orchestrator that receives the underwriting intelligence output and converts it to a loan decision — approve, refer, or decline — with explainability traces for compliance. Handles edge cases: thin files, irregular income, seasonal earnings dips, multi-platform workers. Decisions in seconds, not days.
Compliance Layer
Regulatory infrastructure for NBFC alternative lending operations in India. Covers RBI fair practice codes, KYC pipeline integration, audit trail generation, and explainability requirements for AI-assisted credit decisions. Built so the compliance function can inspect and override without breaking the automated flow.
Ops Automation
Eliminated the manual bottleneck in loan processing. Automated document verification, borrower profile assembly, decisioning handoffs, and exception routing. The ops team now handles exceptions and edge cases — not every application from scratch. Processing capacity scales without proportional headcount growth.
From raw spending data
to a loan decision.
The architecture is a five-stage pipeline. Each stage is isolated, testable, and independently scalable. The output of every stage is a structured artefact — nothing passes between layers as a raw string.
Platform earnings, UPI transactions, wallet activity, cashflow records — pulled via APIs and file exchange. Normalised into a canonical schema.
Spending consistency scores, income stability bands, cashflow volatility index, platform tenure signals, and seasonal adjustment factors computed per borrower.
AI underwriting model ingests extracted features. Outputs a composite risk score with confidence interval and contributing-signal breakdown.
Loan decision engine applies product-level thresholds, bureau-override logic, and compliance guardrails. Produces approve / refer / decline with explainability trace.
Real-time credit outcome with audit trail. Human review queue for edge cases. No bureau score required.
Credit where it was
structurally impossible.
Bureau score required. Gig workers who would be auto-rejected by conventional underwriting are now eligible borrowers — judged on actual financial behaviour.
Loan decisions. Manual review processes measured in days were replaced by an automated pipeline delivering decisions in seconds for standard applications.
Spending-habit signal streams unified into a single risk score. Platform earnings, UPI records, wallet activity, cashflow patterns — one model, one decision.
The system doesn't replace credit judgement. It makes credit judgement possible for a segment of borrowers that the existing infrastructure was incapable of assessing.
Ankur Thakur
CEO, AamdhaneeAamdhanee is a fintech platform providing loans to gig workers in India — underwritten on spending-habit signals rather than conventional credit bureau scores. Non-prime, thin-file lending for the platform economy. aamdhanee.com
Build your AI fintech
infrastructure.
If your lending thesis depends on data that traditional credit infrastructure can't see, that's an engineering problem. We solve engineering problems.
Book a Call