ClearPath Health

🏥 The Problem: A Healthcare System in Crisis

In Colombia, over 700 health lawsuits (tutelas) are filed every single day — mostly because patients can't get timely medical authorizations. Health Insurance Providers (EPS) have accumulated $32.9 billion in debt, and the core bottleneck isn't clinical: it's the slow, manual, administrative auditing process that takes 2+ weeks per case.

Doctors write orders. Patients wait. Bureaucracy kills.

💡 What We Built

ClearPath Health is an enterprise-grade AI agent platform built on UiPath Maestro that automates medical service authorization in seconds. Instead of a human auditor manually reviewing every case, our platform orchestrates a network of specialized AI agents that analyze clinical pertinence, verify EPS coverage, and dispatch a legally valid PDF authorization — all in under 30 seconds.

⚙️ How It Works

The platform runs on UiPath Automation Cloud as the single orchestration and governance layer:

Agent 1 — Clinical Data Reader (UiPath IDP): Extracts patient data, CIE-10 diagnosis codes, and requested procedures from the medical order.
Agent 2 — Coverage Verifier (UiPath Workflows): Queries the EPS benefits registry to validate plan coverage in real time.
Agent 3 — Clinical Analyst (Gemini 2.5 Flash): Evaluates clinical pertinence and calculates a Pertinence Score (0–100).
Agent 4 — Response Generator: Compiles a legal, traceable PDF authorization document.

Decision logic:

Score ≥ 80 → AUTO-APPROVED — PDF generated and dispatched automatically
Score < 80 → HUMAN REVIEW — case escalated to medical auditor via UiPath Action Center

Every case processed is dispatched in real time to the UiPath Orchestrator Queue ClearPathAuthorizations via OAuth2 authenticated REST API.

🔗 Live UiPath Integration (Verified)

This is not a simulation. The integration with UiPath Automation Cloud is real and verified:

OAuth2 Confidential Client authenticated on June 29, 2026
Queue: ClearPathAuthorizations returning HTTP 201 on every submission
Organization: hackathon26_724 · Folder ID: 3087932
Scopes: OR.Queues, OR.Execution, OR.Jobs

🤖 Bonus: Built with AI Coding Agents

This project was built using Claude Code and Cursor as AI coding agents throughout the entire development cycle — from scaffolding the Next.js architecture, to implementing the UiPath OAuth2 integration, to writing the clinical reasoning engine. This qualifies for the UiPath for Coding Agents bonus criteria.

🧗 Challenges We Faced

OAuth2 client_credentials grant required configuring Application Scopes (not User Scopes) in the UiPath External App — a subtle but critical distinction that cost hours of debugging.
Special characters in secrets caused double URL-encoding issues with Python's urllib — solved by switching to requests.
Building a clinical reasoning engine that produces professional-grade medical language without exposing internal scoring math to end users.

📈 Business Impact & Scalability

Eliminates avoidable tutelas from day one
Reduces EPS operational costs dramatically
Human auditors only handle true exceptions — not routine approvals
Multi-tenant SaaS architecture ready to scale: Colombia → Mexico → Brazil → Chile

🛠️ Stack

Next.js 16 · TypeScript · UiPath Automation Cloud · UiPath Orchestrator Queues · Gemini 2.5 Flash · OAuth2 Client Credentials · PDF Generation · Claude Code · Cursor

GitHub: github.com/VIVIANAPLATA16/clearpath-health