FundVista: Mutual Fund Co-Pilot

FundVista is an AI-powered mutual fund dashboard for Indian retail investors. It helps users understand whether they are losing money through Regular mutual fund plans, compare them with Direct plans, analyze portfolio risk, calculate tax and exit-load tradeoffs, and ask a portfolio-aware AI assistant for explanations.

The project is built for a hackathon/demo setting, but the product idea is practical: make hidden mutual fund costs visible and help investors make better, calmer decisions.

---

1. Problem

Non-Technical Explanation

Many mutual fund investors in India do not know the difference between Direct and Regular plans.

Both plans usually invest in the same portfolio and are managed by the same fund manager. The difference is cost:

• Direct plan: lower expense ratio because there is no distributor commission.
• Regular plan: higher expense ratio because commission is paid to an agent/distributor.

This extra cost may look small, like 0.5% to 1.5% per year, but over 10, 20, or 30 years it compounds into lakhs of rupees.

The harder part is that switching is not always obvious. A user must also consider:

• Capital gains tax
• Exit load
• Lock-in period
• Portfolio overlap
• Risk level
• Whether the fund is actually worth holding

Most investors either ignore these details or use multiple scattered tools. FundVista brings them into one decision-making dashboard.

Technical Explanation

The problem is a mix of financial calculation, data modeling, and user experience:

• Store mutual fund data with Direct and Regular variants.
• Track user holdings by session.
• Calculate expense-ratio leakage and future value difference.
• Estimate switching cost using tax and exit-load rules.
• Analyze portfolio health using category, sector, overlap, benchmark, and risk metrics.
• Use AI only after grounding it with portfolio and fund context.

---

2. Solution

Non-Technical Explanation

FundVista acts like a mutual fund co-pilot.

Instead of only showing fund returns, it answers questions like:

• Am I holding Regular plans?
• How much money can I save by switching to Direct?
• Will tax or exit load make switching a bad idea right now?
• Is my portfolio actually diversified?
• Which funds overlap with each other?
• What should I understand before taking action?

The dashboard gives visual answers through charts, calculators, alerts, and an AI assistant.

Technical Explanation

FundVista is a Next.js application with:

• Next.js App Router for frontend and backend routes.
• React + TypeScript for interactive UI components.
• Tailwind CSS + shadcn/ui for a polished dashboard interface.
• Zustand for client-side state management.
• Prisma ORM + SQLite for local structured fund and portfolio data.
• LanceDB vector search for semantic fund retrieval.
• AI chat API that combines conversation history, portfolio context, and retrieved fund context.
• Financial calculators for SIP, SWP, STP, XIRR, tax, exit load, savings, rebalancing, and stress testing.

---

3. Approach

Product Approach

FundVista is designed around one core flow:

Discover funds -> Add holdings -> Analyze portfolio -> Compare Direct vs Regular -> Get switch guidance -> Ask AI for explanation

The goal is not to tell users blindly what to buy. The goal is to make the hidden tradeoffs visible so the user can make an informed decision.

Technical Approach

The app uses a hybrid architecture:

1. Deterministic calculations for finance

   Important numbers such as savings, tax, exit load, SIP future value, XIRR, and portfolio allocation are calculated using code, not guessed by AI.

2. AI for explanation

   The AI Co-Pilot explains the result in simple language. It receives the user's portfolio context and relevant fund data before answering.

3. RAG-style fund context

   The chat route uses semantic search through LanceDB to find relevant fund data, then attaches it to the prompt. This makes answers more specific than a generic chatbot.

4. Session-based portfolio

   Users can try the product without login. Holdings are linked to a session ID and stored locally through the backend database.

5. Modular dashboard

   Features are split into sections:

   • Discover: explore, screener, rankings, heatmap, AMC analysis
   • Analyze: portfolio, comparison, overlap, sector exposure, benchmark, volatility
   • Plan: SIP, SWP, STP, goals, risk profile
   • Optimize: tax, exit load, rebalancing, stress test, alerts
   • Tools: XIRR, watchlist, export, AI Co-Pilot

High-Level Architecture

flowchart TB
    User[User]
    UI[Next.js React UI]
    Store[Zustand Store]
    API[Next.js API Routes]
    Prisma[Prisma ORM]
    DB[(SQLite Database)]
    Vector[LanceDB Vector Search]
    AI[AI Chat / Local Ollama / ZAI fallback]

    User --> UI
    UI --> Store
    Store --> API
    UI --> API
    API --> Prisma
    Prisma --> DB
    API --> Vector
    Vector --> API
    API --> AI
    AI --> API
    API --> UI

Loading

---

4. How To Run

Prerequisites

Install these first:

• Node.js
• npm
• Git

The project already includes a local SQLite database setup.

Step 1: Install Dependencies

npm install

Step 2: Configure Environment

Create a .env file in the project root:

DATABASE_URL="file:./db/custom.db"

Step 3: Generate Prisma Client

npm run db:generate

Step 4: Sync Database Schema

npm run db:push

Step 5: Start The App

npm run dev

Open the app:

http://localhost:3000

Optional: Run Production Build

npm run build
npm run start

Useful Scripts

Command	Purpose
npm run dev	Start development server on port 3000
npm run build	Create production build
npm run start	Start production server
npm run lint	Run ESLint
npm run db:generate	Generate Prisma client
npm run db:push	Push Prisma schema to SQLite
npm run db:migrate	Run Prisma migration
npm run db:reset	Reset database migrations

---

5. What's Next

Product Roadmap

• CAS upload/import: Let users upload Consolidated Account Statements and auto-build their portfolio.
• Live AMFI NAV sync: Keep NAV and fund data updated automatically.
• Broker/platform integrations: Connect with execution platforms for smoother switching.
• PDF portfolio report: Generate a clean advisor-style report.
• Alerts: Notify users about high expense ratio, overlap, manager change, underperformance, or rebalancing need.
• Tax harvesting assistant: Help users plan redemptions more intelligently.

Technical Roadmap

• Add authentication so portfolios persist across devices.
• Replace sample/seeded data with production-grade data pipelines.
• Add stronger test coverage for tax, exit load, and compounding calculations.
• Improve RAG evaluation so AI answers can be measured for correctness.
• Add background jobs for NAV refresh and portfolio alerts.
• Add deployment-ready environment handling for production databases.

---

What Judges Should Notice

FundVista is not just a chatbot and not just a calculator. It combines both:

• Calculators handle exact financial math.
• Database and APIs provide structured fund/portfolio data.
• Vector search retrieves relevant fund context.
• AI explains the result in simple investor-friendly language.
• The interface turns complex decisions into readable dashboards.

In short:

FundVista helps Indian mutual fund investors discover hidden costs, understand risk, and make smarter Direct vs Regular plan decisions with data-backed AI guidance.

---

Tech Stack

Area	Technology
Frontend	Next.js, React, TypeScript
Styling	Tailwind CSS, shadcn/ui, Framer Motion
State	Zustand
Backend	Next.js API Routes
Database	SQLite
ORM	Prisma
AI	AI chat route, Ollama/local fallback, ZAI SDK
Retrieval	LanceDB semantic search
Charts	Recharts

---

Disclaimer

FundVista is an educational and decision-support tool. It does not provide certified financial advice. Users should verify tax and investment decisions with a qualified financial advisor before taking action.