Self-Learning Cheque Processing
- System learns user-verified mappings (e.g., "ABC" on cheque → "BCD Private Ltd" in ledger)
- Achieves 95%+ accuracy in extracting critical data (Party Name, Amount, Date)
- Continuously improves with each verification
Completed
FDC Sales Quotation Bot
Technologies Used
FastAPI LangChain PostgreSQL pgvector OpenAI WhatsApp API Python
Tags
ai automation sales rag whatsapp
Project Links
The FDC Sales Quotation Bot is an intelligent sales automation platform built for an Airia client. It revolutionizes the quotation process by integrating WhatsApp Business API with legacy accounting systems, reducing quotation creation time from 5 minutes to less than 1 minute.
Key Features
RAG Pipeline for Product Identification
- Implemented Retrieval-Augmented Generation using PostgreSQL (pgvector) and OpenAI
- Achieves 99% accuracy in product identification from natural language queries
- Handles complex product catalogs with ease
WhatsApp Integration
- Seamless integration with WhatsApp Business API
- Processes 100+ quotations daily
- Enables 24/7 automated sales operations
Architecture
Backend Stack
- FastAPI: High-performance async Python framework
- LangChain: AI orchestration and chain management
- PostgreSQL + pgvector: Vector similarity search for product matching
- OpenAI: LLM for natural language understanding
Impact
| Metric | Before | After | Impact |
|---|---|---|---|
| Quotation Time | 5 mins | <1 min | 80% reduction |
| Cheque Data Accuracy | Manual | 95%+ | Automated |
| Product Identification | Manual | 99% | AI-powered |
| Daily Processing | Limited | 100+ | 24/7 automated |