MIDDLEGAME · MACHINE LEARNINGMove 08 · Bf4 · 2026
LIVE —:—:—
AI Health & Vehicle Insurance
Fraud Detection & Claim Automation
A proof-of-concept insurance platform combining OCR, serverless AWS compute, and hybrid AI scoring to automate health and vehicle claim verification.
★ Position Gained
Built an end-to-end AI claims pipeline that turns document uploads into real-time fraud verdicts and explainable Truth Scores. The POC demonstrates how rule-based signals and Amazon Bedrock reasoning can reduce manual claim review while improving detection consistency.
Pieces in Play
PythonAWSDynamoDBS3Amazon TextractAmazon BedrockFastAPIReactTailwind CSS
⚔ Complications on the Board
- Designed a hybrid Truth Score engine that balances deterministic fraud rules with Amazon Bedrock AI reasoning.
- Built separate Health and Vehicle claim pipelines while maintaining a shared serverless architecture and UI experience.
- Integrated OCR extraction, document metadata storage, and multi-signal fraud analysis across AWS Lambda, DynamoDB, and S3.
✦ Post-Game Analysis
- Hybrid scoring engines provide both explainability and adaptive fraud detection for insurance claims.
- AgentCore and Nova Pro can augment rule-based workflows with structured AI reasoning for higher-quality verdicts.
- Serverless AWS components make it easier to scale ingest, analysis, and chatbot interactions without managing infrastructure.
Use ← / → to traverse · Esc to return