← Back to work

2026 · Mobile app

SpendSense

Your money, on autopilot.

An AI personal-finance app you barely have to touch. Forward a WhatsApp message, snap a receipt, or just say what you spent, and it logs and categorizes itself. A webhook watches SMS and WhatsApp, a GPT-4o copilot projects your savings (or roasts your habits), and a Smart Spend Radar evaluates impulse buys before you make them.

TypeScriptExpressPrismaPostgreSQLOpenAITesseract.jsTwilioSSE
Backend repo
SpendSense: Your money, on autopilot.
Role
Full-stack: Node/Express + Prisma + Postgres API with OpenAI, OCR, and Twilio, plus the mobile product design
Year
2026
Platform
Android / mobile web

Expense apps fail because logging is tedious. SpendSense removes the typing, capture is passive, and makes the data fun with shareable recaps and social bill-splitting.

What it does

  • Passive logging via a Twilio SMS/WhatsApp webhook, receipt OCR (Tesseract.js), and natural-language or voice entry, parsed by OpenAI
  • GPT-4o financial copilot with tool-calling, with a cheaper GPT-3.5 first pass to keep costs down
  • Smart Spend Radar, a location-aware impulse-purchase evaluator
  • AI anomaly detection (duplicate-charge alerts) and life-event budget prediction
  • Bill-splitting Squads and a gamified Spending Wrapped, updating live over Server-Sent Events

A closer look

Every transaction in one place, splittable with your Squad.
Home: Spending Wrapped, Smart Spend Radar, and AI security alerts.
Log by text, voice, or a receipt screenshot. AI parses it into a transaction.
A webhook listens to SMS/WhatsApp; the radar scores impulse buys by location.
Insights: life-event prediction and an 'invisible expense' anomaly detector.
A GPT-4o copilot that projects savings, finds patterns, or roasts your habits.

Want a walkthrough?

Happy to demo SpendSense live and talk through the decisions behind it.

See more work