Finmentor

A clean and intuitive dashboard showing balance, monthly spending, loan details, and recent transactions at a glance.
Smart search to quickly find users and transfer funds securely
Detailed transaction history with spending categorized by type for smarter financial tracking.
Integrated AI Assistant that helps with queries, transactions, and connects you to agents when needed.
Voice-first financial assistant powered by Groq STT/TTS, making banking as simple as speaking.

Inspiration

FinMentor – Your AI Financial Guide

FinMentor is an AI-powered personal financial assistant designed to make money management simple, smart, and stress-free.
Just like a mentor who guides you in studies, FinMentor acts as a mentor for your finances, helping you understand, plan, and manage your money better.

What does FinMentor do?

It answers your general banking questions like “How to apply for a loan?” or “What’s the interest rate for a savings account?”
It gives personalized investment suggestions, like recommending how much you can invest each month (SIP plans) based on your savings and goals.
It helps with loan management, showing you ways to reduce your repayment burden.
It calculates tax optimization, so you don’t pay extra tax and can claim the right deductions.
It creates simple visual reports of your spending habits, like charts and dashboards, to help you track money easily.

Why is it different?

Unlike a normal chatbot, FinMentor is powered by Generative AI and real-time financial data.
This means it doesn’t just give generic answers – it learns from your profile, your savings, and even official bank resources to provide personalized, accurate, and real-time financial advice.

Who can use it

Anyone looking for a quick, reliable, and 24/7 financial mentor without going to the bank every time.

What it does

FinMentor is an AI-powered financial mentor that:

Provides real-time financial insights with a natural language SQL Agent.
Uses a RAG model with ChromaDB for context-aware answers.
Supports voice-first conversations for hands-free accessibility.
Acts as a tax assistant (salary, business, home-based scenarios).
Includes a loan optimizer to suggest faster repayment strategies.
Generates gamified dashboards and monthly spending charts dynamically.
Supports human agent escalation, bridging AI with real support when needed.

How we built it

Frontend: Custom GPay-style interface with React + Tailwind + shadcn.
Backend: FastAPI + LangChain + Groq Llama 3, integrated with Supabase for user/account data.
Data: Real-time transaction data stored in Supabase and embedded into ChromaDB for RAG.
AI Models: SQL Agent for data queries, Agentic Router for intent classification, and task-specific models for tax, loans, and investments.
Voice: Whisper Turbo for STT and PlayAI TTS for natural-sounding conversations.
Integration: LiveKit for voice streaming and Supabase for authentication + real-time updates.

Challenges we ran into

Designing a modular architecture where each agent (SQL, Tax, Loan, Investment) worked seamlessly.
Handling real-time voice streaming with low latency.
Preventing LLM hallucinations and ensuring precise SQL execution.
Managing agent escalation — we built the trigger system, but connecting live agent calls is still ongoing.
Time management: many late nights debugging and refining features to make the demo smooth.

Accomplishments that we're proud of

Built a full-stack financial assistant from scratch in a short timeline.
Created a SQL Agent that gives accurate financial insights directly from user data.
Designed voice-first interaction with live transcription + AI responses.
Delivered gamified dashboards powered by real transaction data.
Successfully implemented agent escalation trigger as a foundation for hybrid AI-human support.

What we learned

The importance of breaking down a big AI problem into modular agents.
How to combine RAG, SQL, and voice assistants into a single workflow.
Real-world challenges in latency, data accuracy, and scalability.
The value of team collaboration, mentorship, and late-night problem-solving.