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
-
User visits the NBFC website.
-
Chatbot greets user → collects loan requirements.
-
Master Agent triggers:
- Sales Agent → discusses offer
- KYC Agent → fetches CRM data
- Underwriting Agent → runs eligibility logic
-
If approved → PDF sanction letter generated instantly.
-
User receives next steps and feedback summary.
git clone https://github.com/RSN601KRI/lendora-launchpad.git
cd lendora-launchpadnpm installCreate a .env file:
VITE_SUPABASE_URL=
VITE_SUPABASE_ANON_KEY=
OPENAI_API_KEY=
CRM_API_URL=
CREDIT_API_URL=
npm run devThe 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