Calculator0

Start screen with the combination code to enter app
Login Screen
Launch Screen - Safety Plan for user (can be generated my Gemini or manually added)
Upload CSV screen
CSV encryption after upload with industry standard AES
Gemini generated risk assessment (Low risk user)
Custom addition to the safety plan
Resources page

Inspiration

94% of domestic abuse cases involve financial abuse. Financial abuse is the control of another person through financial means. This control is one of the most insidious tactics abusers use to trap their victims. It is also one of the hardest forms of abuse to detect and one of the hardest to escape from. Some of the patterns include the victims wanting to leave but not being able to due to lack of funds. Their paychecks might be intercepted, spending might be monitored, and saving money might be impossible without raising suspicion.

We started with one question: What if victims could secretly analyze their financial situation and plan their escape without their abuser ever knowing?

Most financial tools fail here. Budgeting apps, banking apps, and finance trackers are obvious and easy to monitor.

Then we realized something simple. Everyone has a calculator app. No one questions it.

What it does

Calculator0 is a disguised calculator that hides powerful financial safety tools behind a familiar interface. It is designed to stay invisible while giving users clarity, security, and control.

1. AI-Powered Financial Abuse Detection

Connects securely to bank accounts via Capital One's Nessie API
Uses Gemini 2.5 to analyze 6 months of transaction history
Detects patterns like:
- Missing or irregular deposits (partner intercepting paychecks)
- Sudden spending restrictions (punishment cycles)
- Category elimination (loss of autonomy over time)
- "Allowance" patterns (severe financial control)
Provides risk assessment: LOW, MEDIUM, or HIGH

2. Secret Savings Planner

Calculates how much money victims need to leave safely
Considers: first month's rent, security deposits, emergency fund, moving costs
Tracks progress toward savings goals
All data encrypted and hidden behind biometric authentication
Community resources embedded in the app (shelters, legal aid, hotlines)

3. Encrypted Data Protection

Secure storage for photos, documents, bank statements
AES-256 encryption protects all files
Face ID/email authentication required
Panic button that instantly closes app and clears from recent apps
If discovered, appears as a normal calculator

How we built it

Frontend

React Native + Expo for cross-platform mobile development
Expo Router for navigation
TypeScript for type safety
Expo SecureStore for encrypted local storage

AI & Backend

Google Gemini 2.5 Flash for transaction pattern analysis
Capital One Nessie API for realistic financial data
Python for mock data generation (5 distinct abuse profiles)
AES-256-CBC encryption for sensitive data

AI Prompt Engineering

We crafted specific prompts for Gemini to detect:

Temporal anomalies: Sudden drops after consistent periods
Category diversity: $\text{Autonomy Score} = \frac{\text{Active Categories}}{\text{Expected Categories}} \times 100$
Deposit regularity: $\sigma = \sqrt{\frac{\sum(x_i - \mu)^2}{n}}$ (high variance = red flag)
Amount trends: Declining averages over time

Challenges we ran into

1. Capital One API Transaction Creation

Problem: Even though, we were able to create customers & accounts on The Nessie API but it wasn't actually creating deposits and purchases despite successful POST requests.

Solution: After extensive debugging, we decided to pivot to CSV files. This is still relevant to real world use, as people can download a CSV of account statements from their banking website. We created 5 complete customer profiles with realistic transaction patterns. These files are then encrypted on upload to the app. This actually gave us more control over the abuse patterns we wanted to demonstrate as it allows us to upload the information for each user.

Accomplishments that we're proud of

1. Real-World Impact Potential

We are proud of our work because we didn't just build a tech demo, we built something that could actually create impact for potentially millions of victims of financial abuse. We made our design decision with real survivors in mind: from the disguise to the encryption to the AI patterns. Calculator0 answers real problems that keep people trapped in abusive relationships.

2. Novel AI Application

We believe that using Gemini to detect financial abuse patterns is an innovative use case. We're not just analyzing spending habits, we're detecting coercive control, punishment cycles, and economic isolation. This is pattern recognition that could help millions of people who don't even realize they're being abused.

3. Zero-Knowledge Security Architecture

We implemented a zero-knowledge encryption to ensure the integrity of the critical user information. Every transaction, analysis result, and document is encrypted with the industry acclaimed Advanced Encryption Standard 256 before storage. The user's password never leaves their device, and all decryption happens client-side. This level of privacy is critical for survivors planning a safe exit as well as to protect them from any attack to our platforms.

What we learned

Technical Skills

Expo ecosystem: SecureStore, Constants, Local Authentication, Document Picker
AI prompt engineering: Crafting effective prompts for financial pattern detection
Encryption best practices: AES-256 implementation
API integration: REST APIs, error handling, response parsing
React Native: Cross-platform mobile development, biometric authentication