AI Medical Coding— Glenn Carter

AI medical coding · portfolio

From a clinical note to billable codes — with the receipts.

A working outpatient E/M coding assistant: pick a synthetic note, see the suggested CPT and ICD-10 codes, the exact note span that justifies each one, and the NCCI / documentation flags a real coder would want. Accept, reject, or edit and push back to a (simulated) EHR.

Built by Glenn Carter. Demo runs on hardcoded synthetic notes — no PHI, no live LLM call. The production path (BAA-gated model serving, encryption, audit logging) is on the architecture page.

What this shows

  • Evidence-grounded suggestions. Every suggested code links to the exact span of the note that justified it — not a black box.
  • Real coder workflow. Per-code accept / reject / edit decisions, mirroring the worksheet pattern used in production correction tooling.
  • Rule-engine flags surfaced, not buried. NCCI edits, modifier reminders, LCD warnings, and documentation gaps render alongside the code instead of being hidden in a downstream report.
  • Honest about HIPAA. Architecture page names where PHI flows at every hop and which model-serving paths are BAA-eligible. No “HIPAA compliant” marketing label.
  • Live and deployed. Next.js 16 App Router running at medi.usesmpt.com under TLS, with a documented migration path from this public host to BAA-covered infrastructure for real-PHI evaluation.

For hiring managers

I'm available for remote contract or full-time work on healthcare-AI systems — autonomous coding, clinical NLP, RCM automation, EHR integration. Fastest read on whether I can do the work: open the dashboard, click a chart into the coding view, then skim the architecture page. If that hits the bar, message me at mrglenncarter@yahoo.com.