ClaimPilot Pro

ClaimPilot Pro automates medical billing -- AI reads discharge summary, suggests codes, humans verify, CMS-1500 is generated in seconds and anchored on Polygon for trust.

Comment

ClaimPilot Pro

Inspiration

Every year, billions of dollars are lost to claim delays and coding errors. Hospitals spend days assembling documents for insurers.
We wanted to cut that cycle to minutes while staying compliant and auditable.

What it does

  • Reads clinical discharge summaries or bills.
  • Extracts diagnoses and procedures using AI.
  • Suggests correct ICD-10 and CPT codes with reasoning.
  • Lets billers review, edit, and set claim amounts.
  • Generates a compliant CMS-1500 PDF instantly.

How we built it

  • Frontend: React + Tailwind + shadcn UI for the reviewer dashboard.
  • Backend: FastAPI (Python) micro-services.
  • AI Layer: LLM (Gemini-2.0-flash-exp) for full medical coding;
    optional FAISS semantic search index for ICD/CPT retrieval.
  • OCR: Tesseract pipeline for PDF input.
  • PDF Generation: ReportLab engine for CMS-1500 forms.

Challenges we ran into

  • Balancing automation with human-in-loop verification.
  • Normalizing large ICD/CPT (CPT is licensed, so we created mock dataset) datasets for embedding and search using FAISS index.
  • Ensuring on-chain logging contains no PHI and still remains verifiable.

Accomplishments we’re proud of

  • End-to-end AI coding + human review flow working within 36 hours.
  • CMS-1500 PDF generation fully functional.
  • Cut the average claim preparation time from ~2 days to < 2 minutes in demo tests.

What we learned

  • Real-world health AI needs explainability and compliance, not just accuracy.
  • Human + AI synergy > AI alone.

What’s next for ClaimPilot Pro

  • Fine-tuned medical LLMs for specialty domains with the context of ICD-10 and CPT data access for medical coding.
  • Integrations with EHR systems and clearinghouses.
  • Insurance approval workflow APIs.

Built With

Share this project:

Updates