SpendWise AI

Your AI financial partner for Malaysia. Smart, culturally-aware insights from your transactions, powered by AI. Secure, private, and proactive.

Comment

** SpendWise AI

Inspiration

As Malaysians navigating an increasingly complex financial landscape, we recognized a critical gap: existing financial tools are either too generic or fail to understand our local context. We wanted to build something that truly "gets" Malaysian financial life—from Raya spending to EPF planning. The vision was an AI companion that doesn't just track numbers, but understands the cultural and economic nuances unique to Malaysia.

What it does

SpendWise AI is a comprehensive AI-powered financial management platform built specifically for Malaysia. It combines smart transaction tracking with intelligent insights:

  • Smart Statement Processing: Upload bank statements or receipts, and our Gemini Vision AI automatically extracts transactions with location detection
  • Malaysian-Context Categorization: Automatically categorizes spending as "needs" vs "wants" with local merchant recognition (Shopee, Grab, etc.)
  • AI Financial Assistant: Ask questions like "How can I save more?" and get personalized advice based on your actual spending patterns
  • Credit Card Recommendations: Get personalized card suggestions from Malaysian banks based on your spending habits
  • Cultural Intelligence: Tracks Raya spending, Zakat calculations, school term expenses, and local holiday patterns
  • Privacy-First Design: All PII is masked before AI processing to keep your data secure

How we built it

Backend: FastAPI with Python 3.9+, leveraging Pydantic v2 for validation and SQLAlchemy 2.0 for database operations
Databases: PostgreSQL for transactional data, MongoDB for market/product information
AI/ML: Google Gemini API (Pro and Vision) for intelligent processing, RAG for contextual responses
Architecture: Hybrid database strategy with clear separation between structured transactions and unstructured market data
Security: JWT authentication, bcrypt hashing, PII masking layer before AI processing

Challenges we ran into

  1. PDF Processing Complexity: Malaysian bank statement formats vary widely between banks—we built a flexible extraction pipeline that handles different layouts
  2. Context-Aware Categorization: Teaching AI to distinguish between local merchants and understand Malaysian spending contexts required extensive prompt engineering
  3. Privacy-Utility Balance: Implementing PII masking that protects sensitive data while maintaining AI functionality was challenging
  4. Market Data Integration: Aggregating and structuring Malaysian financial product data from various sources required careful normalization
  5. Real-Time AI Integration: Managing latency and cost while providing near-instant AI responses necessitated smart caching strategies

Accomplishments that we're proud of

✅ Malaysian-First Design: Built a system that truly understands local financial context
✅ End-to-End AI Pipeline: From statement upload to actionable insights in one seamless flow
✅ Production-Ready Architecture: Scalable, secure system ready for real users
✅ Innovative RAG Implementation: Context-aware AI that combines user data with market intelligence
✅ Complex Problem Solving: Successfully processed and categorized transactions from 15+ Malaysian bank formats
✅ Privacy Innovation: Developed proprietary PII masking that maintains AI utility while protecting user data

What we learned

  • Local Context Matters: Financial tools must understand cultural and regional nuances to be truly useful
  • AI + Rules = Better Results: Pure AI categorization has limitations; hybrid approaches with rule-based logic work best
  • Privacy is Non-Negotiable: Users will only trust financial AI if they're confident their data is protected
  • Malaysian Financial Literacy: There's a huge appetite for accessible financial guidance in Malaysia
  • Multi-Language Processing: Support for both English and Bahasa Malaysia significantly increases accessibility
  • Market Gaps: Many Malaysians are using multiple disconnected tools—there's clear demand for an integrated solution

What's next for SpendWise AI

Short-term (Next 3 months):

  • Mobile app development (React Native)
  • Enhanced budgeting tools with envelope system
  • Integration with Malaysian tax planning

Medium-term (Next 6 months):

  • Open banking API integrations with Malaysian banks
  • Family/shared account features
  • Investment portfolio tracking

Long-term (Next 12 months):

  • SME financial management features
  • Insurance product marketplace
  • Retirement planning with EPF/PRS integration
  • Expansion to other Southeast Asian markets

SpendWise AI represents more than just a technical achievement—it's our commitment to empowering Malaysians with financial intelligence that understands their world. We're building the financial companion we wish we had, and we're just getting started.

Share this project:

Updates