Landing Page
Create Account Page
Login Page
Bank Syncing Page
Transactions Syncing Page
Dashboard Insights Pt. 1
Dashboard Insights Pt. 2
Dashboard Insights Pt. 3

CapitalClarity

“Smart spending starts with smart awareness.”

CapitalClarity is a Gemini AI-powered React Native app designed to help users take control of their finances. By analyzing your spending history, the app delivers real-time, personalized financial insights that empower you to make smarter, budget-conscious decisions. Think of it as your intelligent financial coach, always learning and guiding you.

The Problem

Many people—especially students—struggle with managing personal finances. The issue isn’t always about income, but rather a lack of awareness and context around spending.

“I check my bank app and just think… where did all my money go?”

In fact:

42% of college students carry credit card debt, often unaware of how daily habits accumulate.
Small, frequent purchases tend to slip under the radar, making budgeting difficult.
Existing financial tools often feel disconnected or require too much manual input.

Inspiration

CapitalClarity was born out of personal pain points:

“We’ve both had moments where we blew our entire paycheck in a weekend—without really knowing how.”

On-campus job earnings would quickly disappear with little to show for it.
Impulse spending became a norm due to lack of financial feedback in real-time.
We needed something proactive—not reactive—that could learn from us and help build healthier habits.

The Project

CapitalClarity doesn't just track expenses—it understands them. By combining AI-powered reasoning with user-specific context, the app classifies expenses, flags potentially wasteful spending, and adapts over time.

Technologies Used

Tech	Why We Chose It
React Native	Unified mobile development across platforms, enabling quick iteration and a polished user experience.
Flask (Python)	Gave us access to Python’s ecosystem for processing data and running AI models efficiently.
Gemini AI 2.5	Handled deep analysis—e.g., determining if a purchase at "Target" was an emergency or impulse buy.
Gemini AI 2.0 Multi-Input	Allowed fast, continuous input handling for adaptive learning based on real-time transaction streams.

What We Learned

Financial behavior is complex – Classifying spending as “unnecessary” often requires understanding personal context, not just transaction data.
User-centric design is crucial – The more intuitive the interface, the more likely users will stick around and benefit.
End-to-end integration matters – Building a fast, reliable pipeline from Plaid → Backend → AI → UI was key to delivering real-time insights.
Feedback loops improve accuracy – Our predictions became much sharper once we added in user feedback and historical behavior reinforcement.

“The model got better once it listened to what we had to say.”

⚠️ Challenges Faced

1. Real-Time Data Pipeline

Getting real banking data is hard. We attempted to use the Plaid API to seamlessley integrate with Capital One, Chase, and other bank accounts, but were blocked by multi-day api approval.

Thus, we had to pivot into creative solutions for getting transaction data, emulating the real-time nature we wanted the application to capture.

2. Accurate Predictions with Limited Data

Here’s the tough part: we often had only the vendor name and amount for each transaction.

“$8.99 at Taco Bell —was it a meal or a craving induced snack”

To tackle this, we:

Analyzed time of transaction to detect patterns
Used merchant categories to infer context (e.g., groceries vs. fast food)
Collected user feedback to fine-tune prediction weights

This created a feedback loop that made the app smarter the more you used it.

Looking Ahead

CapitalClarity is still evolving. Some ideas we’re excited to explore:

Better AI Insights to teach users about their spending habits in a personal manner
Bank Integrations to allow for a more seamless approach to our application
Personalizing the app, recognizing different cultures and spending habits to create a more inclusive environments.