FinSight AI — Smart Personal Finance Advisor

Upload your bank statement CSV and get instant AI-powered spending analysis, smart insights, and a personal finance chatbot. No login. No bank connection. Just upload and understand your money.

Comment

Inspiration

As students, we never really knew where our money was going. Existing finance apps are overwhelming, require bank logins, and give generic advice. We wanted something dead simple — upload a CSV, get instant answers. No signup, no bank connection, no complexity.

What it does

FinSight AI analyzes your bank statement CSV and gives you:

  • Spending breakdown by category with pie and bar charts
  • AI-generated insights — warnings, saving tips, and positive habits
  • Top merchants by total spend
  • Spending narrative — a plain-English summary of your month
  • Finance chatbot — ask anything like "Am I spending too much on food?" or "How can I save ₹5000 more?"

How we built it

  • Frontend: React (Vite) + Pure CSS, Recharts for charts, PapaParse for CSV parsing
  • Backend: Node.js + Express REST API
  • AI: Groq API with LLaMA 3.3 70B Versatile — chosen for its speed and accuracy
  • Deployment: Vercel (frontend) + Render (backend)

The flow is simple: CSV → PapaParse → Express backend → Groq LLaMA 3.3 70B → structured JSON → React dashboard.

Challenges we ran into

  • Getting Groq to return consistent structured JSON every time required careful prompt engineering with strict output format instructions
  • Handling diverse CSV formats from different banks — column names vary wildly
  • Deploying frontend and backend as separate repos while keeping CORS and API routing working correctly

Accomplishments that we're proud of

  • Built and deployed a full-stack AI product end-to-end
  • The chatbot gives genuinely useful, context-aware financial advice — not generic responses
  • Clean, polished dark UI that feels like a real product

What we learned

  • Structuring API responses as strict JSON requires careful prompt design — output format instructions make a huge difference
  • Groq's LLaMA 3.3 70B is surprisingly fast and accurate for financial categorization tasks
  • Separating frontend and backend repos adds complexity but makes deployment and scaling much cleaner
  • Building a full-stack AI product end-to-end in a single session taught me more than weeks of tutorials

What's next for FinSight AI — Smart Personal Finance Advisor

  • Support for multiple months of data with trend analysis over time
  • WhatsApp bot integration so users can ask finance questions via chat
  • Multi-currency support beyond INR
  • Budget goal setting — set a monthly limit per category and get alerts
Share this project:

Updates