What We Plan to Build

We are building FinMentor, an Agentic AI-powered Financial Mentor that delivers personalized, real-time financial guidance using a multi-agent system.
Unlike conventional chatbots, FinMentor uses specialized autonomous agents that collaborate to understand, reason, and act intelligently on user queries.

Each agent focuses on a distinct financial domain, while a Central Agentic Router dynamically coordinates their interactions to deliver accurate, contextual, and human-like responses.

Core AI Agents in FinMentor:

🧠 Conversational Agent – Handles user interaction via text/voice using Groq Whisper Turbo for speech-to-text and LLaMA 3 (70B) for reasoning.
🧾 SQL Agent – Converts natural language into SQL queries to fetch live financial insights from Supabase, offering transparent financial summaries.
💸 Loan Agent – Suggests optimized repayment structures, refinancing opportunities, and prepayment benefits.
💰 Investment Agent – Provides personalized SIP, mutual fund, and savings recommendations based on user transaction patterns.
💼 Tax Agent – Calculates estimated tax and identifies deductible expenses using embedded financial rules and historic transaction data.
🧩 RAG Agent – Uses Agentic RAG (Retrieval-Augmented Generation) with ChromaDB to retrieve user-specific context and enhance factual accuracy.
🧠 Query Optimization Agent – Refines vague user inputs into precise, data-driven prompts, improving output clarity and model performance.
🗣 Voice Agent – Enables real-time voice interaction with speech-to-text and text-to-speech flow.
👨‍💼 Human Escalation Agent – In emergencies, triggers connection to a live human financial advisor for secure and personalized help.

The result is a collaborative, intelligent ecosystem that acts as a 24/7 personal financial coach, capable of continuous learning, reasoning, and autonomous task execution.

How the Agentic System Works

Input Understanding: Voice or text input captured and transcribed by Whisper Turbo.

Intent Analysis & Query Enhancement: The Query Optimization Agent enhances user queries for higher accuracy.

Agentic Routing: The Agentic Router assigns the query to the relevant agent (SQL, Tax, Loan, or Investment).

Context Retrieval: The RAG Agent fetches contextual data from ChromaDB and Supabase for grounded reasoning.

LLM Reasoning: The LLaMA 3 model generates personalized insights based on contextualized data.

Response Delivery: The Conversational or Voice Agent delivers the output as text or speech.

Through this intelligent pipeline, FinMentor transforms raw queries into meaningful financial actions.

Pain Points Addressed

Reactive vs Proactive Systems: Existing financial tools track spending but don’t guide or coach users.

Irregular Income Challenges: Gig workers and freelancers struggle to maintain savings and manage unpredictable cash flows.

Fragmented Financial Platforms: No single ecosystem integrating tax, investment, and loan planning seamlessly.

Generic Insights: Current AI tools give broad suggestions; FinMentor delivers context-aware, data-grounded recommendations.

Lack of Voice & Emergency Support: Traditional chatbots can’t handle spoken inputs or escalate issues to human agents.

FinMentor’s agentic design bridges these gaps — for example, the Loan Agent collaborates with the SQL Agent to fetch repayment data before advising restructuring options.

Target Audience

Gig workers, freelancers, and informal sector professionals with irregular income.

Young professionals seeking smart saving, investment, and tax insights.

Banks and financial institutions aiming to embed personalized financial coaching into their apps for better customer satisfaction and data security.

Fintech startups that can leverage modular AI agents through APIs.

Proof of Concept (POC)

A fully functional prototype has been developed as a web application, where users can log in, manage transactions, and interact via text or voice.
All transactions are securely stored in Supabase, enabling the SQL Agent and RAG Agent to generate real-time insights, tax summaries, and visual dashboards.

This POC demonstrates the end-to-end capability of agentic reasoning, multi-agent communication, and personalized decision-making — making financial mentorship both intuitive and intelligent.

Go-To-Market (GTM) Strategy & Revenue Streams

B2B Integration with Banks & Fintechs:

Integrate FinMentor’s agents (SQL, Tax, Loan, Investment) into financial apps to provide personalized experiences.
Ensures higher customer satisfaction and data privacy, as data remains within the bank’s infrastructure.

Freemium SaaS Model for Individuals:

Free Tier: Access to dashboards and summaries.

Premium Tier: Includes advanced agents (Tax, Loan, Investment) and voice features.

AI-as-a-Service (API Model):

Offer individual agents — Query Optimizer, RAG Engine, SQL Agent — as APIs for third-party fintech developers.