Our Solution – DhanMitra

1. An AI-powered financial coaching agent that:

• Learns from real financial data (via Fi Money MCP + user input).

• Tracks spending & income behavior continuously.

• Detects risks before they hurt.

• Gives personalized, actionable recommendations in simple language.

• Works offline (local LLM) for privacy + real-time response, and online (cloud LLM) for deeper insights.

How It Works :

1. Data Layer: MCP server, bank feeds, investment data, market info.

2. Hybrid AI Layer:

• Local Gemma (fast, private) → drafts analysis, detects habits.

• Cloud LLM (powerful, deep) → validates insights, adds intelligence when online.

• Dual-pass reasoning: If online, cloud verifies; if offline, local model self-critiques.

3. Agent Layer:

• Budget Cop 🛑 – prevents overspending.

• Risk Radar ⚠️ – detects income/expense mismatches.

• Wealth Builder 📈 – nudges toward SIPs, FDs, investments.

• Debt Doctor 💊 – monitors EMIs, loans, credit health.

4. User Layer:

• Chat interface → advice in natural conversation.

• Notifications/reminders → proactive coaching

Demo Flow

1. User connects MCP → DhanMitra ingests spending & income.

2. Local Gemma flags: “You spent 40% more on food this month.”

3. Cloud model enriches (if online): “At this pace, your balance will dip ₹5,000 before month-end. Consider reducing by ₹200/day.”

4. User asks: “How much should I save for my upcoming trip?”
   → Agent calculates from current cashflow + gives plan.

5. Nudges: “Transfer ₹2,000 to your SIP tomorrow – it aligns with your monthly pattern.”

Why DhanMitra is Unique

✅ Personalized coaching (not generic tips).

✅ Hybrid AI → Offline-first + Cloud-smart.

✅ Multiple specialized agents working together.

✅ Secure by design (sensitive data stays local).

✅ Culturally relevant brand – “DhanMitra” feels like a trusted friend.

Impact

• Helps gig workers, freelancers, and students manage irregular cashflows.

• Reduces financial stress → better savings + smarter investments.

• Builds financial literacy passively (user learns as they interact).

• Scales to robo-advisory, MSME financial health, and even micro-credit risk analysis.