MediLink Healthcare Companion

Insurance Claim Dashboard
Main Dashboard

Inspiration

This project was inspired by the everyday struggles many people in the United States face when trying to navigate healthcare, insurance, and medical bills. We observed patterns across friends, family, and news stories: people unsure if symptoms warrant an emergency room visit, confusion during appointment booking, and a maze of insurance denials and surprise bills. These issues are compounded by broader financial pressures, such as: Total Financial Stress = Healthcare Costs Housing} + Credit Card Debt + Student Loans

Keywords AI Organization Name & ID

Name: Hackathon ID: e320f9a8-b456-4874-9f4a-c2f075472dfd

We wanted to build an AI-powered advocate to help patients make safer decisions, stay organized, and fight back against a system that often says “no” by default.

What it does

MediLink acts as both a navigator and an advocate through five specialized AI agents:

Symptom Triage Chatbot: Helps users determine if they should monitor at home, book an appointment, or seek urgent/emergency care.

Smart Appointment Assistant: Manages appointments through a unified interface, interacting with provider systems via APIs or structured workflows.

Medication & Visit Tracker: Syncs visits and follow-ups while providing medication reminders.

Post-Discharge Follow-up Agent: Proactively checks in after hospitalization (e.g., "Did you take your antibiotics?") and surfaces red-flag symptoms.

Insurance & Billing Advocate: Matches plans to in-network providers, explains denial letters in plain language, identifies billing errors, and automates structured appeal letters.

How we built it

The system utilizes a modular architecture to remain extensible and maintainable:

Frontend: Built with Lovable for rapid UI development and responsive workflow orchestration.

Backend: Implemented with FastAPI (Python) to expose RESTful endpoints for triage, booking, and claim analysis.

AI Layer: Leverages GPT-4o for clinical-style reasoning support and parsing complex insurance denial letters.

Orchestration: Managed via Keywords AI to break complex workflows into multi-step agents and trace reasoning over time.

Database & Auth: Handled by Supabase for storing user profiles, medication schedules, and generated appeal artifacts.

Infrastructure: Deployed on AWS with a focus on encryption, role-based access, and audit logging toward HIPAA-aligned controls.

Challenges we ran into

Safety and Scope: We had to carefully constrain the system to support decision-making without replacing a doctor, emphasizing "triage and education" over "diagnosis".

Insurance Complexity: Policies vary significantly by payer, plan, and state. Representing these variations for AI reasoning while keeping the UX simple was non-trivial.

Privacy: Handling sensitive health data required treating security as a first-class feature rather than an add-on during the hackathon.

Workflow Integration: Many providers still rely on legacy systems like fax or semi-structured PDFs, requiring creative workflows like generating structured letters for manual upload.

Accomplishments that we're proud of

What we learned

Administrative vs. Medical: The largest pain points are often administrative—understanding insurance vocabulary and catching billing errors—rather than just medical confusion.

Agentic Power: Multi-step AI processes are exceptionally powerful for cross-referencing policy language with claim details and assembling personalized arguments for appeals.

AI Guardrails: Building safe healthcare AI requires clear escalation pathways to human clinicians and official guidelines.

What's next for MediLink Healthcare Companion

Billing Advocacy Expansion: Introducing specialized agents to optimize insurance claims and further reduce administrative overhead for patients.

Intelligent Care Pathways: Ensuring even deeper collaboration between agents to guide patients to the best holistic healthcare options.

Advanced Privacy: Future enhancements will include advanced encryption, continuous HIPAA compliance improvements, and comprehensive audit trails for all data access.