RAGnarok
Team consisting of a Rockwell SWE-2, a Zensar AI Engineer, and a ChiStats Data Scientist skilled in Java, Python, Flutter, RAG, and microservices.
YouTube Video
Project Description
The Problem MedAuth AI Solves
The Crisis in American Healthcare Right Now
Prior Authorization (PA) is the process where a doctor must get insurance company approval before a patient can receive a medication or procedure. It sounds reasonable - until you see how it actually works:
What Happens Today (Without MedAuth AI)
Doctor decides patient needs Humira for Crohn’s Disease
Staff pulls up Blue Cross IL’s PA form- a 12-page PDF
They manually fill patient info, diagnosis codes, drug dosage, prior therapy history
They fax it (yes, in 2026, still fax) to the insurance company
They wait 3-14 days for a decision
Often it’s denied - wrong form, missing field, wrong ICD-10 code
Patient waits in pain. Sometimes weeks.
If denied, staff starts an appeal - another form, another fax, another wait
Every insurer has different forms. Every drug has different criteria. Every diagnosis has different rules. That’s thousands of unique combinations - and doctors deal with dozens per week.
Why No Static App Can Fix This
Every combination of payer × drug × diagnosis is a completely different form:
Scenario
What’s Required
Humira + Crohn’s + Blue Cross IL
Step therapy proof (methotrexate, 6-MP), TB test, endoscopy report
Humira + Crohn’s + Aetna
All of the above + gastroenterologist attestation + board cert + 6-month clinical notes
Keytruda + Lung Cancer + UnitedHealth
Completely different — PD-L1 score, EGFR/ALK testing, tumor board minutes, oncologist NPI
That’s 3 completely different applications for the same drug and same diagnosis. A static form either shows everything (unusable) or nothing useful.
Doctor types: “Humira for John Doe, 45M, Crohn’s, failed methotrexate, Blue Cross IL”
↓ (1.6 seconds)
AI Agent:
→ Parses: patient, drug, payer, clinical history
→ Looks up: Blue Cross IL’s exact PA requirements
→ Checks: which step therapy criteria are already met
→ Renders: the EXACT right form, pre-filled, with only the fields that apply
→ Flags: what’s missing before you even submit
→ On denial: generates a medically-worded appeal letter instantly
The doctor’s staff goes from 45+ minutes of manual form-filling to reviewing a pre-filled, correct form in under 2 minutes.
The Generative UI Insight
This is the core innovation — the UI itself is the output of the AI agent.
Not just answers. Not just text. The agent generates:
The right form fields (different for every payer)
The right sections (Aetna gets a gastro attestation section; Blue Cross doesn’t)
The right document upload slots (exactly which supporting docs are needed)
The right clinical criteria checklist (auto-checked from the input)
No static app can do this. You’d need to hardcode thousands of form variations. The only solution is an AI that generates the UI dynamically based on context