An intelligent personal finance assistant powered by AI — helping users track spending, detect fraud, simulate transactions, and receive personalized financial insights in real time.

Financial illiteracy is a widespread problem, especially among younger generations. According to the National Financial Educators Council, the average U.S. adult lost $1,819 in 2022 due to poor financial knowledge. Meanwhile, a 2023 FINRA survey revealed that nearly two-thirds of Americans cannot pass a basic financial literacy test.

The problem is systemic — most young adults don’t receive proper financial education until they are already facing real-world responsibilities like budgeting, taxes, or debt. To make things worse, hiring a professional accountant or financial advisor is often unaffordable, with rates ranging from $150 to $400/hour.

We built this app to offer an accessible, 24/7 AI-powered alternative — an accountant you can carry in your pocket.

This project serves as a personal AI accountant and financial advisor that follows the user anywhere. Users can chat with an intelligent agent that:

• Analyzes past transactions by category
• Simulates sending money to other users
• Detects potential fraud based on transaction patterns
• Offers personalized financial insights using age, interests, and user history
• Retrieves current financial news and trends using Retrieval-Augmented Generation (RAG)

• Frontend: Next.js – for a responsive UI and chat experience
• Backend: Flask – to handle AI logic and DB interactions
• Database: Supabase – user profiles, transactions, wallets
• AI Agent: LangChain + Gemmini (2.0 Flash)
• Vector Search: Pinecone – powers RAG for trend insights
• Embedding & Retrieval: Retrieval-Augmented Generation (RAG)

• LangChain Learning Curve: Our first time building tools, memory, and multi-tool agents
• RAG Integration: Wrapping our heads around vectorization and Pinecone took time
• Pinecone Setup: Authentication and document retrieval setup was tricky
• Tool Chaining: Ensuring the agent could reason about and properly use tools with live data
• Framework Overload: Many of the frameworks (LangChain, Supabase, Pinecone) were new to us

• Created a real AI agent that behaves like a financial assistant
• Integrated GPT, LangChain, and Supabase for intelligent decision-making
• Simulated real financial workflows (like sending money) through backend logic
• Used RAG to deliver up-to-date insights about financial spending
• Delivered a polished, chat-based interface to interact with AI tools

• How to implement multi-tool LangChain agents with memory and reasoning
• The fundamentals of vector databases and how RAG works
• Prompt engineering to guide agent behavior with real user data
• Secure and efficient financial logic across a full-stack architecture
• Cross-system integration and agent design in a real-world context

• Add savings goals, budgeting tools, and financial reminders
• Expand RAG to learn from user-specific historical patterns
• Add better fraud detection models using AI-based anomaly detection
• Integrate calendar-based financial summaries and planning
• Polish and deploy for real users to test and iterate on feedback

Coming soon — once deployed, you’ll be able to test the agent, send money, analyze spending, and receive insights directly from your dashboard!

Want to collaborate, contribute, or learn more? Feel free to reach out to the team or submit an issue!