An AI-driven, conversational, end-to-end personal loan sales assistant designed for NBFCs. This project leverages Agentic AI architecture, enabling seamless automation across customer engagement, KYC verification, credit evaluation and instant sanction-letter generation — all through a web-based chatbot.

Traditional NBFC loan journeys are slow, manual, and impersonal. Customers face long verification steps, unclear eligibility rules, and generic offers — leading to low digital conversion rates.

Our solution introduces an Agentic AI Loan Sales Assistant that replicates a human sales officer but operates with the speed, accuracy, and transparency of AI.

✔ Conversational & personalized

✔ Automated KYC & credit checks

✔ Real-time underwriting logic

✔ Instant PDF sanction letter generation

✔ Explainable & auditable decisions

• Master Agent: Handles conversation, identifies intent and orchestrates tasks.

• Worker Agents:

  • Sales Agent – loan discussion & offer negotiation
  • KYC Agent – validates user details from mock CRM
  • Underwriting Agent – evaluates credit score & eligibility
  • Sanction Letter Agent – generates a PDF instantly

Built using React + Tailwind + shadcn/ui, providing:

• Smooth chat experience
• Dynamic prompts
• Real-time decisioning

• Node.js / Python-based APIs
• Credit Score API (mock)
• CRM API (mock)
• AutoML-enabled scoring logic

lendora-launchpad/
│
├── public/                 # Static assets
├── src/
│   ├── components/         # UI components (chat UI, inputs, layouts)
│   ├── agents/             # Master & Worker AI Agents
│   ├── hooks/              # Reusable logic
│   ├── lib/                # Utilities, configs
│   ├── pages/              # Page-level UI
│   ├── services/           # APIs (CRM, Credit Score, Underwriting logic)
│   └── types/              # Typescript interfaces
│
├── supabase/               # DB config (if using Supabase)
│
├── index.html
├── package.json
├── vite.config.ts
└── README.md               

• React + TypeScript
• Tailwind CSS
• shadcn/ui
• Vite

• LangChain
• GPT-based Worker Agents
• Node.js / Python
• PDFKit / ReportLab (PDF generation)

• Supabase / PostgreSQL

• Deployed on Vercel / AWS

1. User visits the NBFC website.

2. Chatbot greets user → collects loan requirements.

3. Master Agent triggers:

   • Sales Agent → discusses offer
   • KYC Agent → fetches CRM data
   • Underwriting Agent → runs eligibility logic

4. If approved → PDF sanction letter generated instantly.

5. User receives next steps and feedback summary.

git clone https://github.com/RSN601KRI/lendora-launchpad.git
cd lendora-launchpad

npm install

Create a .env file:

VITE_SUPABASE_URL=
VITE_SUPABASE_ANON_KEY=
OPENAI_API_KEY=
CRM_API_URL=
CREDIT_API_URL=

npm run dev

The system follows a modular, scalable Agentic Orchestration Architecture with clear separations between:

• Conversation Layer
• Intelligence Layer
• Decision Layer
• Data Layer
• Output Generation Layer

✔ 25–30% increase in conversion rate

✔ Loan decisions in < 10 minutes

✔ 30% reduction in operational effort

✔ Improved CSAT & trust through explainable AI

✔ Scalable across geographies and loan products

• Multilingual agent support
• Voice-enabled interactions
• Federated learning for secure model improvement
• Adaptive emotional intelligence modelling

• GitHub Repository
• Demo Link
• Figma Wireframes: https://www.figma.com/board/Hp6zEyCsIR6OeC7FZT9KOM/FinGenie-UX-Flow-Diagram-- Customer-Journey-?node-id=0-1&p=f
• Architecture PDF from EY submission

Algoric Team – EY Techathon 6.0 Finalists

• Aryan Panda – AI, Fullstack, Workflow Design
• Roshni Kumari – Data Science, ML, Feature Engineering