MedPass AI is a comprehensive, modern healthcare workflow platform designed to bridge the gap between hospitals, insurers, and patients. It automates insurance approvals, analyzes hospital discharge blockers in real-time, and strictly adjudicates claims without hallucination, avoiding existing manual verification processes and speeding up evaluation, approvals and clarity for the stake holders.
In addition to operational workflows, the platform includes Trace Commons—an isolated, strictly-governed database engine that converts live hospital data into 100% anonymous, highly-monetizable operational datasets for researchers and underwriters, fully compliant with the DPDP Act 2023.
- Hospital Administration Portal: Easily track active admissions, view real-time discharge blockers, and monitor the pre-authorization pipeline through beautiful, semantic data visualizations.
- Insurer Adjudication Portal: A secure, dedicated environment to review claims, automatically map billing codes, and instantly flag fraudulent or missing data via a deterministic rules engine.
- Patient Portal: A jargon-free, transparent view of a patient's insurance coverage, showing exactly what is covered, what was rejected, and why.
- Context-Aware Assistance: Ask our built-in assistant about cases, claim statuses, and policy details.
- Deterministic Guardrails: The chatbot seamlessly reads clinical data and returns beautifully formatted, accurate markdown answers. We implemented strict word-boundary NLP matching to completely eliminate patient-name hallucinations during generic queries.
- Zero-PII Architecture: Strictly reserved for administrators, this engine scrubs all Protected Health Information (PHI) and Personally Identifiable Information (PII). Exact dates become generalized months, and exact ages become 15-year brackets.
- Premium Data Formats: Generates research-ready ZIP bundles containing enriched CSVs and highly-compressed Apache Parquet files, including detailed Data Dictionaries mapping clinical and financial domains.
| Hackathon Track | Implementation Details |
|---|---|
| 1. GitHub Developer Track | Implemented a clean Git tree, robust GitHub Actions CI/CD pipeline (automatically builds frontend and validates backend syntax on push), CODEOWNERS, and issue/PR templates. |
| 2. Beeceptor Mocking Track | Built a mock client & simulator for Insurer Preauthorization, NHCX Claim Submission, and FHIR endpoints handling Success, Query, Rejection, and 500 error scenarios. |
| 3. Render Deployment Track | Multi-service blueprints, containerized backend configurations, and highly-optimized Vite/React frontend deployed via Render with strict environment variable configuration. |
| 4. n8n Automation Track | Engineered a decoupled outbound webhook dispatcher firing on case lifecycle events (e.g., CLAIM_SUBMITTED, STATUS_CHANGED) with a ready-to-import n8n operational workflow. |
flowchart TB
classDef secure fill:#e0f2fe,stroke:#0369a1,stroke-width:2px;
classDef op fill:#f0fdf4,stroke:#15803d,stroke-width:2px;
classDef trace fill:#fdf4ff,stroke:#a21caf,stroke-width:2px;
classDef external fill:#f3f4f6,stroke:#374151,stroke-width:2px;
classDef extBox fill:#f9fafb,stroke:#d1d5db,stroke-width:1px,stroke-dasharray: 5 5;
%% STAKEHOLDERS
subgraph Stakeholders ["Users & Stakeholders"]
H[🏥 Hospitals]
I[🛡️ Insurers]
P[🧑⚕️ Patients]
R[🔬 Researchers]
end
%% EXTERNAL
subgraph Ext ["External Integrations"]
EMR[Hospital EMR/FHIR]:::external
TPA[Insurer / TPA API]:::external
NHCX[NHCX India]:::external
end
%% MEDPASS OPERATIONAL
subgraph MedPass ["MedPass AI Platform (Operational)"]
direction TB
subgraph Frontend ["Frontend (Web)"]
Dash[Role-Based Dashboards<br/>React + TypeScript]
Viz[Charts & Analytics]
end
subgraph Backend ["Backend Services (FastAPI)"]
CE[Case & State Engine]
PE[Policy & Coverage Engine]
Calc[Financial Waterfall]
DI[Discharge Intelligence]
Doc[Document OCR Parsing]
end
subgraph Database ["Operational Database"]
PG[(Supabase PostgreSQL)]
Auth[Auth & Privacy RLS]
end
end
%% TRACE COMMONS
subgraph TraceCommons ["Trace Commons (Data Layer)"]
direction LR
Ingest[Event Ingestion]:::trace --> Std[FHIR Standardization]:::trace
Std --> DeID[Presidio De-identification]:::trace
DeID --> DQ[Data Quality]:::trace
DQ --> Storage[(DuckDB + Parquet)]:::trace
Storage --> Export[CSV / Parquet Export]:::trace
end
%% FLOWS
H & I & P -- "Secure Access (JWT)" --> Dash
Dash --> Backend
Backend <--> PG
Backend <--> Auth
Backend <--> Ext
Backend -- "Transactional Outbox Events" --> Ingest
R -. "Access De-identified Data" .-> Export
Export --> ML[Future ML Models]:::trace
Export --> Analytics[Research Analytics]:::trace
class Stakeholders,Dash,Viz,CE,PE,Calc,DI,Doc,PG,Auth secure;
MedPass AI follows a B2B Healthcare SaaS + AI Analytics business model with multiple revenue streams across hospitals, insurers, and healthcare intelligence.
| 💰 Revenue Stream | 🎯 Description |
|---|---|
| 🏥 Hospital SaaS Subscription | Subscription plans for hospitals to manage insurance and discharge workflows. |
| 🧾 Per-Case Transaction Fees | Charges for each insurance pre-authorization or claim processed. |
| 🏢 Enterprise Hospital Plans | Custom deployments for hospital chains and healthcare networks. |
| 🛡️ Insurer / TPA Integration | API integrations for automated claim verification and adjudication. |
| 🎨 White-Label Solutions | Custom-branded MedPass AI platform for hospitals and TPAs. |
| 🔌 API & Integration Services | FHIR, OCR, webhook, and insurance APIs for third-party systems. |
| 📊 Analytics & Intelligence | Premium dashboards for operational, financial, and claim analytics. |
| 🔬 Trace Commons Data Products | Anonymous healthcare datasets for research and enterprise analytics. |
| 🤖 AI Automation Add-ons | OCR, chatbot, discharge summaries, and workflow automation modules. |
| 🧠 Research & Healthcare Intelligence | Longitudinal disease insights and population health analytics via Trace Commons. |
Revenue Strategy: SaaS subscriptions • Per-case processing • API licensing • Enterprise plans • AI automation • Privacy-preserving analytics.
A modular, API-first stack for insurance authorization, discharge intelligence, and governed healthcare data.
| Layer | Technology | Purpose in MedPass AI | Role |
|---|---|---|---|
| Frontend | React + TypeScript | Role-based stakeholder dashboards | Hospital • Insurer/TPA • Patient • Trace Commons |
| UI System | Tailwind CSS + shadcn/ui | Consistent, accessible interface components | Reusable cards, tables, dialogs and workflow UI |
| Visualization | Recharts / ECharts | Operational and Trace analytics | Case funnel, readiness, cohort and trend views |
| Backend API | FastAPI + Python | REST APIs and domain orchestration | Cases, claims, documents, Trace, integrations |
| ORM / Data Access | SQLAlchemy | Typed persistence and domain data access | Operational schema + Trace schema |
| Operational DB | PostgreSQL (Supabase) | System of record for operational workflows | Cases, policies, claims, documents, audit and organizations |
| Auth & Access | Supabase Auth + JWT / RLS | Authentication and tenant-aware access control | Target production security model |
| Policy & Decision | Deterministic Python engines | Coverage, calculation, decision and readiness logic | Rules first; evidence-backed explanations |
| Document Intelligence | OCR + LLM assistance | Extract structured fields from medical/insurance documents | LLM assists ambiguity/extraction; not the authoritative decision maker |
| Event Layer | Transactional Outbox | Reliable domain-event capture | Admission → diagnosis → treatment → authorization → discharge → outcome |
| Data Layer | Canonical schema | Governed longitudinal healthcare representation | Privacy-preserving subject/facility/encounter and clinical/workflow domains |
| Privacy | Presidio + deterministic rules | PII detection, de-identification and policy gates | Fail-closed publication gate; configurable governance |
| Analytics / Export | DuckDB + Parquet | Cohort queries and efficient dataset exports | Timeline, hospital and cohort-based selection |
| Interoperability | FHIR + OMOP adapters | Exchange and research-standard representations | FHIR for interoperability; OMOP for research analytics |
| Data Quality | Validation framework | Completeness, temporal and referential checks | Quality report attached to dataset versions |
| Lineage / Provenance | OpenLineage-compatible model | Transformation and dataset provenance | Track source → transformation → published version |
| Mock Integrations | Beeceptor | Simulate payer/external APIs | Hackathon-friendly integration testing |
| Workflow Automation | n8n (optional) | Non-core workflow automation | Use only where automation adds value |
| Container / Runtime | Docker | Reproducible application packaging | Backend/frontend deployment consistency |
| CI/CD | GitHub Actions | Automated build, test and deployment checks | Quality gate for every change |
| Deployment | Render / Vercel | Host web apps, APIs and database | Cloud deployment path for MVP/demo |
MedPass AI is designed to create measurable impact across the entire healthcare insurance ecosystem—from patients and hospitals to insurers and research organizations. Instead of improving just one workflow, it streamlines the complete insurance authorization, discharge, and healthcare data lifecycle.
| 👥 Stakeholders | 🎯 Primary Impact |
|---|---|
| 🧑⚕️ Patients | Transparent insurance journey and financial clarity |
| 🏥 Hospitals | Faster discharge coordination and reduced manual work |
| 🛡️ Insurers / TPAs | Structured claims with fewer review delays |
| 🔬 Research & Analytics | Anonymous longitudinal healthcare datasets through Trace Commons |
Empowering patients with transparency throughout the insurance and discharge process.
| Impact Area | Benefit |
|---|---|
| 🛡️ Insurance Transparency | View real-time insurance approval, rejection, and pending claim status. |
| 💰 Financial Visibility | Understand covered amount, co-pay, deductible, and remaining payable amount before discharge. |
| 📄 Reduced Document Requests | Avoid repeatedly submitting the same insurance or hospital documents. |
| 📋 Actionable Guidance | Receive clear next steps for pending approvals, missing documents, or insurer queries. |
| ✅ Discharge Readiness | Know exactly what is blocking discharge and what actions remain. |
Reducing operational burden and improving insurance workflow efficiency.
| Operational Challenge | MedPass AI Impact |
|---|---|
| 📑 Manual Insurance Coordination | Automates insurance verification, claim tracking, and communication workflows. |
| Instantly identifies incomplete medical or billing information before submission. | |
| 💵 Patient Liability Visibility | Clearly separates insurer-payable and patient-payable charges. |
| 🚦 Case Prioritization | Highlights urgent pre-authorizations, discharge blockers, and pending insurer actions. |
| 📊 Auditability | Maintains structured audit trails for every insurance workflow and decision. |
Improving claim quality and enabling faster adjudication.
| Workflow Improvement | Benefit |
|---|---|
| 📂 Cleaner Claim Submissions | Standardized medical bills, discharge summaries, and supporting evidence. |
| 🧾 Structured Evidence Extraction | OCR + AI converts unstructured documents into machine-readable claim data. |
| ❓ Fewer Avoidable Queries | Detects missing fields before claims reach insurer review. |
| ⚙️ Consistent Review Workflows | Deterministic policy engine applies the same validation rules for every case. |
| 📈 Operational Analytics | Monitor approval rates, rejection reasons, turnaround time, and workflow bottlenecks. |
| 📍 Impact Dimension | 🚀 Outcome |
|---|---|
| Patient Experience | Transparent insurance journey with fewer delays. |
| Hospital Operations | Reduced manual coordination and faster discharge processing. |
| Insurance Processing | Higher-quality claims and faster adjudication workflows. |
| Healthcare Data Intelligence | Anonymous longitudinal datasets for research and analytics through Trace Commons. |
This project includes a fully automated Continuous Integration & Continuous Deployment (CI/CD) pipeline via GitHub Actions.
- Continuous Integration (CI): On every push or pull request to the
mainbranch, GitHub Actions automatically provisions an Ubuntu environment, installs all Node.js/Python dependencies, builds the Vite frontend, and validates the backend Python syntax. - Continuous Deployment (CD):
- Frontend: Deploys instantly via Vercel. Any updates to the
mainbranch trigger a live production build. - Backend: Deploys seamlessly to cloud platforms (like Render or Railway) by dynamically binding to the
$PORTenvironment variable.
- Frontend: Deploys instantly via Vercel. Any updates to the
cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Seed the database with demo cases and trace encounters
python ../scripts/seed_demo.py
python ../scripts/seed_trace.py
# Start the server
uvicorn app.main:app --reload --port 8000cd frontend
npm install
npm run devNavigate to http://localhost:5173 in your browser.
- Typography Overhaul: Replaced external CDNs with Plus Jakarta Sans (Google Sans) for a crisp, enterprise-grade UI aesthetic.
- Dynamic Infographics Redesign: Overhauled the Hospital Preauth Pipeline and Discharge Blocker charts with live data mapping, semantic colors, and smooth rendering animations.
- Trace Export Expansion: Upgraded the Trace Commons engine to produce richly detailed clinical/financial columns without compromising PII constraints.
Built with ❤️ for DSU DevHack 3.0