CoinCise

What Inspired Me

Money is one of the most emotional parts of life, yet most finance apps treat it like a spreadsheet problem.
I noticed that people around me (including myself) didn’t struggle because they couldn’t calculate, but because they didn’t understand why they spent, what it meant, and what it silently affected.

Most tools answer:

• Where did my money go?

But almost none answer:

• Why did I spend this?
• What did this choice replace?
• Will I regret this later?

What I Learned

While building this project, I learned several important things:

1. AI is most powerful when invisible
   The best use of AI wasn’t a flashy chatbot, but quiet reasoning behind features like semantic receipts, intent detection, and decision reflection.

2. Behavioral finance matters more than formulas
   Users don’t need complex equations upfront — they need clarity, reassurance, and context.
   Math supports decisions, but psychology drives them.

3. Constraints improve design
   By limiting the AI to specific roles (teacher, challenger, explainer), the app became clearer, safer, and more useful.

4. Good UX reduces financial anxiety
   Calm language, friendly names, and soft insights made the app feel supportive instead of judgmental.

Architecture Overview

CoinCise is built as a simple, clean, modular system: Here are the features of the project; some are not included due to lack of time but will be added to enhance the functionality. The overall tech stack planned for the project is mentioned below; some functionalities are to be added soon in the future.

• Frontend (Next.js + React)
  Handles the user experience: expense input, feature blocks, insight cards, and learning views.

• Backend (Cloudflare Workers)
  Acts as a secure, serverless brain that:

  • Receives user actions
  • Applies business logic
  • Communicates with AI and the database

• AI Layer (Gemini API)
  Used for:

  • Expense intent classification
    
  • Semantic receipt generation
    
  • Finance concept explanations
    
  • Investment education
    
  • Decision counter-arguments
    

• Database (MongoDB Atlas)
  Stores:

  • User expenses
    
  • Spending patterns
    
  • Learning progress
    
  • Financial personality profile
    

Data Flow

User Action
   ↓
Backend
   ↓
Gemini AI (Reasoning & Teaching)
   ↓
Insight returned to User

Built With

• css3
• gemini
• javascript
• mongodb
• node.js
• react