From the project archive
AI-Powered FinTech B2B Invoice Management
Predicting payment delays with ML
- My role
- Full Stack Developer
- Timeline
- 4 months
- Year
- 2022
- Format
- Project
- Prediction Accuracy
- 89%
- Invoices Processed
- 10K+
- Response Time
- <200ms
Context
B2B finance teams spend countless hours manually triaging invoices to prioritize collection efforts. There was no data-driven way to know which customers would pay late.
Approach
Combined a React frontend and Java backend for invoice CRUD operations with a Python ML service that scores each invoice's likelihood of on-time payment using historical features.
The build
A unified dashboard where ops teams can filter invoices by risk score, drill into predicted payment dates, and export reports — all backed by a scalable PostgreSQL schema.
Outcome
Reduced manual invoice-triage time by an estimated 60% in simulated runs and surfaced high-risk accounts two weeks earlier than heuristics.