PiggyBank is a real-time financial behavior agent that helps users make smarter spending decisions before they happen.
Instead of tracking what you already spent, PiggyBank predicts what you’re about to spend — and nudges you at the exact moment it matters.
Not “you spent too much” 👉 but “you’re about to spend — here’s a better choice”
Backend deployment target: Railway Frontend deployment target: Vercel
Railway backend:
- Start command:
uvicorn backend.main:app --host 0.0.0.0 --port $PORT - Health check path:
/healthz - Readiness check path:
/readyz - Required env vars:
DATABASE__URL,TWILIO__ACCOUNT_SID,TWILIO__AUTH_TOKEN,TWILIO__PHONE_NUMBER,GEMINI__API_KEY,PLAIDCLIENT_ID,PLAIDSECRET - Browser allowlist env var:
APP__CORS_ORIGINS=https://your-frontend.vercel.app
Vercel frontend:
- Required env var:
VITE_API_BASE_URL=https://your-backend.up.railway.app - Production must point at the Railway backend URL; the frontend no longer falls back to localhost.
Most financial stress doesn’t come from big purchases.
It comes from:
- $6 coffees
- $15 food deliveries
- small, repeated habits
These “invisible expenses”:
- quietly drain your money
- reduce financial confidence
- go unnoticed in the moment
PiggyBank is a messaging-first AI system that:
- Learns your spending habits
- Predicts upcoming purchases
- Nudges you before you spend
- Connects daily decisions to long-term goals
PiggyBank works like texting a smart friend.
User texts Piggy:
coffee 6.50
Piggy responds:
Oink 🐷 Got it — logged your coffee!
Piggy sends:
You're likely to grab coffee soon 👀
Skip today → you're $6 closer to your $250 bike
User replies:
maybe 😭
Piggy adapts and responds intelligently.
- SMS input (primary)
- Optional receipt image (OCR)
Users can:
- type purchases manually
- send receipt photos for automatic extraction
PiggyBank models spending behavior using:
- time between purchases
- time-of-day patterns
- frequency & consistency
- recency-weighted behavior
It outputs:
- predicted purchase window
- probability of purchase
- confidence score
Using Gemini, Piggy decides:
- whether to send a nudge
- when to send it
- how strong the message should be
Piggy communicates via SMS:
- proactive nudges
- goal-based motivation
- behavioral feedback
- financial insights
User (SMS)
↓
Twilio Webhook
↓
Backend (FastAPI)
↓
Prediction Engine
↓
Gemini Decision Layer
↓
Twilio Response (SMS)
↓
User
↓
Feedback Loop → Model Updates
- Messaging: Twilio (SMS)
- Frontend (Dashboard): Next.js (mobile-first web)
- Backend: FastAPI (Python)
- Database: PostgreSQL
- AI Layer: Gemini API
- Optional: OCR for receipt processing
PiggyBank is built mobile-first, but not as a traditional app.
👉 Primary interface: SMS (real-time, zero friction) 👉 Secondary interface: lightweight web dashboard
This makes the experience:
- immediate
- natural
- integrated into real life
- 📊 Predictive spending detection
- 💬 Real-time SMS nudges
- 🧠 Adaptive behavior learning
- 🎯 Goal-based motivation
- 🧾 Receipt parsing (optional)
- 🤖 Conversational financial assistant
Most finance apps are reactive:
“Here’s what you spent”
PiggyBank is proactive:
“Here’s what you’re about to spend — and what to do instead”
Piggy continuously learns from user behavior:
- Did the user follow the nudge?
- Did they ignore it?
- Did they respond?
This updates:
- prediction accuracy
- messaging tone
- intervention timing
- Bank integration via Plaid
- Advanced ML prediction models
- Reinforcement learning for nudging
- Personalized financial insights dashboard
Built by a 2-person team focused on:
- behavioral AI
- real-time systems
- human-centered design
PiggyBank isn’t just a budgeting tool.
It’s a real-time behavioral feedback system that turns small daily decisions into meaningful financial progress.