Smarter insights. Better habits. Stronger finances.

FinSight

Inspiration

When meeting new people, we often use frameworks like MBTI to quickly understand how they think and behave. This idea inspired us to ask: what if we could understand financial behavior the same way?

Today, most people don’t struggle with accessing financial data — they struggle with interpreting it. Bank statements are cluttered, overwhelming, and lack meaningful guidance. As a result, many users overspend without fully understanding why.

We wanted to bridge that gap by transforming raw transaction data into something intuitive, behavioral, and actionable.

What We Built

FinSight Assistant is an AI-powered fintech application that analyzes transaction data to uncover spending patterns, identify behavioral tendencies, and provide actionable recommendations.

Instead of simply tracking expenses, FinSight:

Classifies a user’s financial personality
Explains the why behind spending habits
Suggests targeted interventions to improve financial behavior

For example, if a user frequently makes small purchases, FinSight may identify a high-frequency spending pattern and recommend introducing a delay before purchases or setting a weekly cap.

How We Built It

We built FinSight using a modern full-stack architecture:

Frontend: React + Vite for a fast, responsive interface
Backend: Vercel serverless functions to securely handle API requests
AI Engine: Google Gemini API for natural language insights
Styling: TailwindCSS for clean, minimal UI design

We designed a pipeline where:

Users input transaction data
The backend processes and formats the data
Gemini generates insights based on behavioral patterns
The frontend presents results in a clear, user-friendly format

Mathematically, we think of spending patterns as a function:

$$ Behavior = f(Transaction\ Frequency,\ Category\ Distribution,\ Spending\ Variance) $$

This allows us to map raw financial data to interpretable behavioral outputs.

What We Learned

Through this project, we learned:

How to integrate AI models into real-world applications
The importance of secure API handling (moving from frontend to backend)
How to design systems that prioritize user understanding over raw data
The value of combining behavioral psychology with technology

We also gained experience working with serverless architectures and deploying scalable applications using Vercel.

Challenges We Faced

One of the biggest challenges was ensuring that sensitive API keys were not exposed. Initially, our AI calls were handled on the frontend, which posed a security risk. We resolved this by moving all AI interactions to a backend serverless function.

Another challenge was structuring AI outputs in a way that felt useful and actionable rather than generic. This required careful prompt design and iteration to ensure the responses were meaningful.

We also had to align branding and user experience across the app to ensure consistency and clarity, especially for a live demo environment.