AI-Powered Underwriting System
Enterprise underwriting platform built for corporate insurance workflows, with C1 (my team) focused on the AI intelligence layer for OCR, decision support, quotation generation, and site risk analysis.
Project Overview
This project delivered a web-based corporate underwriting system designed to automate key parts of the traditional underwriting workflow. The C1 scope focused on the AI intelligence layer, covering document data extraction, underwriting recommendation support, quotation generation, and site intelligence. The system was designed to reduce manual data entry, improve consistency, and support underwriters with structured decision assistance rather than replacing human judgment.
Key Highlights
Built the AI intelligence layer covering OCR extraction, underwriting decision support, quotation generation, and site risk analysis.
Implemented an OCR pipeline using Tesseract with template-based extraction and OpenCV-based checkbox detection for proposal forms.
Designed a rule-based referral engine with critical escalation rules and non-critical premium adjustment rules.
Developed quotation generation logic that applies configurable risk multipliers based on construction, occupancy, and security inputs.
Integrated site intelligence using geocoding, weather data, satellite imagery, and AI-based visual risk scoring.
Delivered a four-stage AI-assisted workflow while keeping the underwriter in control of the final decision.
Problem
Traditional underwriting workflows rely heavily on manual data entry, document verification, and subjective risk evaluation. These processes create inefficiencies, increase turnaround time, and lead to inconsistent outcomes. A major challenge was to introduce intelligent automation while preserving human oversight in a regulated insurance workflow.
Solution
Team C1 implementation delivered a four-stage AI-assisted pipeline. First, an OCR module extracts proposal form fields using Tesseract, coordinate-based templates, and OpenCV-based checkbox detection. Next, a rule-based referral engine evaluates critical underwriting rules for escalation and non-critical rules for premium adjustment. A quotation generation model then applies pricing multipliers to produce structured premium outputs. In parallel, a site intelligence module resolves addresses, retrieves weather data and satellite imagery, and applies AI-based scoring across flood, construction, roof, fire, and weather risk dimensions.
Outcome
The project demonstrated a working AI-assisted underwriting MVP where OCR, business-rule evaluation, quotation generation, and site intelligence were integrated into a unified workflow. The result was a system that reduced manual effort, improved decision consistency, and showed how external AI and geospatial services can be embedded into an enterprise underwriting process while keeping the underwriter as the final decision-maker.