Insurance
Enterprise LLM Platform for Insurance Underwriting
Engineered an enterprise-scale LLM platform processing 10,000+ insurance documents daily, slashing underwriter review time by 97%.
Key Outcomes
- — 97% reduction in underwriter review time
- — 10,000+ documents processed daily
- — 90%+ extraction accuracy
The Challenge
Insurance underwriting requires extracting accurate, structured data from thousands of complex, unstructured documents daily — applications, medical records, financial statements. Manual processing was the bottleneck, and the stakes for errors are high: regulatory exposure, missed risk signals, bad pricing.
The Approach
As Head of AI at Sixfold, I engineered an enterprise-scale LLM platform purpose-built for insurance document processing, with compliance and accuracy as first-order concerns.
Key Contributions
- Engineered enterprise-scale LLM platform processing 10,000+ insurance documents daily, slashing underwriter review time by 97%
- Formulated comprehensive PII/PHI protection protocols that exceeded HIPAA and SOC 2 Type 2 compliance requirements
- Developed proprietary LLM evaluation framework that slashed hallucination rates, ensuring extraction reliability at scale
- Created customized document analysis workbench, increasing extraction accuracy to above 90%
The Impact
The platform transformed underwriting from a document-heavy manual process into an AI-assisted workflow where underwriters focus on judgment calls rather than data entry — with the compliance guardrails to operate in a regulated industry.