VTdevelops/AI-Powered-Underwriting-Risk-Copilot

AI-Powered Underwriting Risk Copilot

An interactive Gradio application that assists insurance underwriters by extracting structured insights from submission documents, summarizing key risk exposures, detecting anomalies, scoring risks, and drafting underwriting notes.

Features

• —Upload PDF submissions or paste submission text/email content.
• —Extract structured data points such as insured name, line of business, payroll, TIV, and loss history.
• —Summarize risk exposures leveraging an LLM-friendly pipeline with rule-based fallbacks.
• —Flag anomalies like missing values, high loss ratios, or inconsistent metrics.
• —Generate an explainable risk score and a draft underwriting note that underwriters can edit.

Local Development

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python app.py

The Gradio interface launches on http://127.0.0.1:7860.

Hugging Face Spaces Deployment

1. 1.Space Type: choose Gradio (Python).
2. 2.Upload all source files (app.py, requirements.txt, and the src/ package).
3. 3.Spaces automatically install requirements and launch the Gradio interface defined in app.py.
4. 4.Add any required environment variables (for hosted LLM APIs) in the Space settings.

Project Layout

.
├── app.py                 # Gradio entrypoint
├── requirements.txt       # Python dependencies
└── src
    └── risk_copilot
        ├── __init__.py
        ├── config.py
        ├── ingestion.py
        ├── extraction.py
        ├── scoring.py
        ├── summaries.py
        └── notes.py

Next Steps

• —Integrate organization-specific underwriting guidelines for scoring and anomaly rules.
• —Connect to historical policy/loss databases for richer analytics.
• —Capture underwriter feedback to continuously refine extraction and scoring logic.