claims-denials

Agentic denial-triage for medical billing teams. Reads EOBs and remittance advice with GPT-5.6, routes each denial to the correct resolution path — appeal, corrected claim, patient bill, or write-off — and drafts the artifact for human review. Built with Codex for OpenAI Build Week 2026.

Revenue cycle analytics platform for denial prevention, appeal prioritization, payer friction, underpayment discovery, work queues, ML scoring, and GitHub Pages deployment.

Flags healthcare claims likely to deny before submission, cutting simulated rework cost 38.5% on held-out future claims. Leakage-safe DuckDB SQL feature layer, calibrated sklearn models compared on a time split, cost-based decision threshold, batch CLI scoring 45k claims/sec. Synthetic data only, no PHI.