Assets Feature
Dashboard
Chat Feature

WealthScope - Hackathon Submission

Inspiration

The inspiration came from a personal frustration: having wealth scattered across multiple platforms without being able to see the complete picture. Stocks at Interactive Brokers, crypto at Binance, a rental apartment, bonds at the bank, and physical gold stored away. Every month, getting a real balance sheet required opening 5 different apps and maintaining a manual Excel spreadsheet.

We thought: "Why doesn't a Mint or Personal Capital exist that includes ALL assets, not just bank accounts? And why not use AI to actually advise me?"

That's how WealthScope was born: your personal CFO in your pocket that sees ALL your wealth and tells you what to do with it.

What it does

WealthScope is a mobile app (Flutter) that consolidates all your assets into one intelligent dashboard:

🎯 Unified Dashboard

Aggregates stocks, bonds, crypto, real estate, gold, commodities
Calculates your REAL net worth including illiquid assets
Visualizes portfolio distribution with interactive charts

🤖 Conversational AI (Google Gemini)

24/7 financial chat: "Should I sell my apartment to buy stocks?"
Predictive analysis: "What if I sell X and buy Y?"
Contextualized responses to your real financial situation

📸 Smart OCR

Scan real estate contracts, property titles, gold certificates
Automatically extract amounts, dates, terms
Create assets in the system without manual entry

💰 Shadow Pricing

Values illiquid assets without market price (real estate, physical gold)
Uses comparables and alternative metrics with AI
Keeps your wealth automatically updated

📊 What-If Analysis

Simulate scenarios: "What if inflation rises 5%?"
Compare investment strategies
Make data-driven decisions, not emotional ones

🌅 Daily Briefings

Personalized morning summary of your financial situation
Market alerts relevant to YOUR portfolio
Daily action recommendations

How we built it

Tech Stack

Backend (Go 1.23 + Gin)

Clean Architecture: isolated domain, easy to test
GORM for PostgreSQL (Supabase)
Swagger/OpenAPI for automatic documentation
RevenueCat webhooks for payments
Yahoo Finance API for market prices

Frontend (Flutter 3.x)

Riverpod for state management
Feature-first architecture
fl_chart for visualizations
flutter_secure_storage for tokens
image_picker + OCR for document scanning

Artificial Intelligence

Google Gemini 3.0 for conversational analysis
Gemini Vision for financial document OCR
Specialized prompts for financial analysis

Infrastructure

Backend on Railway (auto-deploy from GitHub)
Database on Supabase (PostgreSQL with RLS)
Auth with Supabase Auth (JWT)
RevenueCat for paywall and subscriptions

Development Process

Week 1: Foundations

Project architecture (backend + frontend)
Authentication system with Supabase
Data model for multi-type assets
Basic dashboard with charts
Asset CRUD (CREATE, READ, UPDATE, DELETE)

Week 2: AI Features

Google Gemini integration
Conversational chat with financial context
OCR for document scanning
Shadow pricing for illiquid assets
What-if analysis with simulations

Week 3: Polish & Demo

Automatic daily briefings
Paywall with RevenueCat
End-to-end testing
Complete documentation
Demo video and presentation

Git Workflow

We used simplified GitFlow:

main - stable production
develop - continuous integration
feature/* - new features
PRs with code review before merge

Challenges we ran into

1. Heterogeneous Asset Modeling

The biggest technical challenge was designing a data model that supports stocks, real estate, gold, bonds, and crypto in the same table without losing flexibility.

Solution: We used a hybrid model with common fields (asset_type, quantity, current_value) and a JSONB metadata field for type-specific fields. Example:

{
  "ticker": "AAPL",
  "exchange": "NASDAQ"
}

2. Shadow Pricing of Illiquid Assets

How do you value an apartment you bought 2 years ago without paying for an appraisal every day?

Solution: We created a "shadow pricing" system that:

Uses reference indices (price/sqft in the area)
Applies estimated depreciation or appreciation
Allows manual updates when there's a real appraisal
Uses Gemini to suggest values based on comparables

3. Context Window in Gemini

Gemini has context limits. If a user has 100 assets, how do we send everything in each question?

Solution: We implemented "relevance filtering":

Only send the top 10 most relevant assets (by value)
Add summarized metrics (total value, % per category)
The rest is included only if the user asks specifically

4. Market API Handling

Yahoo Finance is free but has rate limits and sometimes fails.

Solution:

Price caching system (valid for 15 minutes)
Fallback to previous day's prices if API fails
api_quotas table to monitor usage and avoid blocks

5. Onboarding UX

Asking the user to add 10 assets manually in the first session is a guaranteed "drop-off".

Solution:

OCR to scan documents (quick entry)
Simplified manual import (just name + value)
Progressive onboarding: "Add 1 asset now, the rest later"
Demo mode with example portfolio

6. Testing Non-Deterministic AI

How do you test a feature where Gemini can give different answers?

Solution:

Integration tests verify response structure, not exact content
Mocks for unit tests
Manual tests with predefined prompts to validate coherence

Accomplishments that we're proud of

🏆 Functional MVP in 3 Weeks

From zero to a functional app with Go backend + Flutter frontend + AI in record time. It's not a prototype, it's an app we would actually use.

🎨 Polished UX

Despite limited time, the app feels professional:

Smooth animations
Native dark mode
Interactive charts
Elegant error handling

🤖 AI that Actually Helps

It's not a generic chatbot. Gemini has context of YOUR portfolio and gives personalized advice. You ask "Should I sell my Tesla shares?" and it responds based on how many you have, what they're worth, and your risk profile.

📷 Functional OCR

The ability to scan a purchase agreement and have an asset created automatically is magical for the user. It eliminates massive friction.

🏗️ Scalable Architecture

Clean Architecture in backend means we can:

Switch from PostgreSQL to MongoDB without touching business logic
Test without database (mocks)
Add new asset types without massive refactors

📚 Complete Documentation

Wiki with 10+ technical documents:

Product definition
System architecture
API design
Data model
Sprint plan
Security & privacy

Anyone can enter the project and understand what to do.

🔒 Security-First

From day 1:

Authentication with Supabase (JWT)
Row Level Security in database
Input validation on all endpoints
No secrets in code (env vars)
HTTPS only

What we learned

Technical

1. Google Gemini is Amazing for Finance

Understands complex financial concepts without fine-tuning
Gemini Vision for OCR is superior to Tesseract
Long context handling (1M tokens) is a game-changer

2. Flutter is Ideal for Rapid Prototyping

Hot reload = 10x development speed
Single codebase for iOS + Android
Riverpod makes state management predictable

3. Clean Architecture is Worth It

Although it takes more time initially, it pays dividends
Tests are trivial (easy mocks)
Requirement changes don't break everything

4. Supabase is Magic

Auth, database, and RLS out-of-the-box
Open-source alternative to Firebase
"Real" PostgreSQL vs limited NoSQL

Product

1. Onboarding is Everything

If you ask for 10 assets manually, the user leaves
OCR and demo mode reduce friction massively
"Add later" > "Complete now"

2. AI Needs Real Context

Generic ChatGPT doesn't work for personal finance
The value is in connecting AI with YOUR data
Personalized responses > general advice

3. Mobile-First is Right for Finance

People check finances on their phone (Uber, coffee, etc.)
Push notifications = engagement
Desktop would be a secondary feature

Team

1. Early Documentation Accelerates

Wiki from day 1 = less Slack, more code
API design upfront = backend + frontend in parallel
Clear sprint plan = everyone knows what to do

2. Simple Git Flow Works

You don't need full Gitflow for 2 people
Feature branches + PRs = enough
CI/CD from week 1 avoids "works on my machine"

3. Scope Creep is Real

We had to say "no" to 10+ cool features
MVP = Minimum, not Maximum
"After the hackathon" is a valid phrase

What's next for WealthScope

Short Term (Post-Hackathon)

🚀 Closed Beta Launch

100 early adopter users
Weekly feedback loop
Iterate on UX based on real usage

📊 More Data Sources

Integration with brokers (Alpaca API, Interactive Brokers)
Import from CSV/Excel
Bank connections (Plaid/Belvo)

🔔 Proactive Notifications

"Tesla dropped 5% today, you have $5k at risk"
"It's a good time to rebalance your portfolio"
Target price alerts

Medium Term (6 months)

🤖 More Sophisticated AI

Fine-tuning Gemini with proprietary financial data
News sentiment analysis
Price prediction (with disclaimers)

📈 Auto-Trading (Premium)

Execute trades based on AI recommendations
Automatic portfolio rebalancing
Automated DCA (Dollar Cost Averaging)

👥 Social Features

Share investment strategies
Performance leaderboards (anonymous)
Collaborative investment groups

🌍 International Expansion

Multi-currency support
European and Asian markets
Localization (English, Portuguese, Spanish LATAM)

Long Term (1-2 years)

🏦 Become a Neo-Bank

Integrated investment accounts
Debit cards linked to portfolio
Asset-collateralized loans

🎓 Financial Education

In-app investment courses
Market simulator (paper trading)
Certifications and badges

🤝 B2B for Financial Advisors

Dashboard for advisors with multiple clients
White-label for financial institutions
Public API for developers

🔐 Decentralization (Web3)

Integrated crypto wallet
DeFi positions tracking
NFTs as assets

Final Vision

WealthScope is not just a finance app. It's the financial copilot we all deserve.

In 5 years, we want people to say: "Before WealthScope, I didn't really know how much I was worth or what to do with my money. Now I have clarity and confidence."

Democratizing premium financial advisory is not just a pitch, it's our mission.

Team: 2 developers, 3 weeks, 1 dream.
Stack: Go + Flutter + Gemini + PostgreSQL
Status: MVP ready for beta testing
Next Step: Turn this hackathon into a real product that changes lives.

🚀 Let's build the future of personal finance, together.