The situation
Insurers, energy advisers and students all work from dense documents: claim forms, bills and long PDFs. Reading them by hand is slow and error-prone.
AI chatbots with RAG
Three RAG chatbots that turn insurance claims, electricity bills and long PDFs into accurate, grounded answers.
Insurance, energy and education
01 · Context
02 · Challenge
What had to be true
Each chatbot had to find the right facts in the source documents and answer precisely, not guess.
Projects
Three chatbots
01
Medical CPT code extraction from insurance claims
A RAG chatbot that extracts medical CPT codes from insurance documents and answers claim questions in context. An LLM reads scanned claim forms and maps medical terms to codes, backed by a database of insurance terminology and codes for real-time lookup. Claim types are classified automatically and coding errors flagged.
02
Information extraction from electricity bills
For a marketing company helping customers move to lower-cost electricity providers. The chatbot extracts usage, unit costs and total charges from bills, compares providers and their pricing plans, and Python logic recommends the best option for each customer's usage. Automated notifications tell users about potential savings.
03
Q&A bot over PDF documents
Built for workshop demonstrations at VIT Chennai and Amrita University, Coimbatore. The chatbot indexes large PDFs, retrieves the most relevant passages and re-ranks them with an LLM to answer academic, legal and technical questions accurately.
03 · What we built
The system
- Document ingestion with OCR for scanned forms and bills
- Indexing in vector databases and SQL for real-time lookup
- Retrieval with LLM re-ranking to surface the most relevant passages
- Domain logic on top: code mapping and error checks for claims, cost comparison and savings alerts for bills
04 Architecture
How it works, step by step.
Step through the system, or let it play.
- Claims, bills or PDFs arrive
- OCR and chunking
- Embed and index
- Retrieve and re-rank
- LLM answers with context
- Codes, savings or answers returned
05 · Outcome
Insurers and healthcare providers cut manual effort in claim processing while keeping accuracy, and energy customers see their savings without reading a bill.
06 · Stack
Next case · HR · Gen AIInterview agent using LLMs