Real problems, the architecture behind the solution, and what it changed.

AI-Powered Insurance Policy Intelligence Platform

Reads insurance policy documents and returns structured, classified benefit data through an API.

policies processed per month

5,000+

policies processed per month

extraction time per policy (about 93% less)

4 h → 15 min

extraction time per policy (about 93% less)

accuracy in benefit identification and classification

~95%

accuracy in benefit identification and classification

• Python
• FastAPI
• Azure OpenAI
• LangChain
• Azure AI Search

RAG-Based Benefit Matching Engine

Aligns extracted benefits to a standard catalogue using semantic retrieval and fuzzy matching.

of catalogue alignment automated

~85%

of catalogue alignment automated

• Python
• LangChain
• ChromaDB
• Hugging Face embeddings
• Sentence Transformers

HBB File Transformation Engine

Turns raw, inconsistent benefit data files into industry-standard Excel output automatically.

• Python
• Pandas
• Regex
• OpenPyXL
• Excel automation

AI Code Auditor — Multi-Agent Engineering System

A multi-agent architecture for auditing a code repository: quality, security, architecture and reporting.

• Python
• LLM tool calling
• Agent orchestration
• AST / static analysis tools
• Git

Modern Agentic AI Applications

An ongoing set of agentic AI experiments: MCP, tool calling, multi-agent collaboration and human-in-the-loop design.

• Python
• Model Context Protocol (MCP)
• LangGraph
• LangChain
• AWS Bedrock AgentCore