CareLoop – Healthcare Access Through AI Voice

Inspiration

Healthcare access should not depend on owning a smartphone, having internet connectivity, or understanding complex digital applications. Many patients, especially elderly individuals and people in rural communities, struggle with appointment management, follow-ups, and accessing clinic information.

We wanted to build a solution that makes healthcare coordination as simple as making a phone call.

What We Built

CareLoop is an AI-powered healthcare coordination assistant that enables patients to interact with clinics through natural voice conversations. It supports Hindi-first conversations with automatic switching between Hindi, English, and Hinglish.

Patients can:

Book appointments through voice conversations.
Check appointment details and availability.
Reschedule or cancel appointments with explicit confirmation.
Receive automated appointment reminders and follow-up calls.
Initiate missed-call callbacks without requiring internet access.
Access clinic information and emergency guidance.

Clinicians receive a dedicated workspace to manage appointments, supervise follow-up calls, and monitor call statuses.

How We Built It

We developed CareLoop using a modular architecture consisting of:

Frontend: React, Vite, and Tailwind CSS.
Backend: Node.js and Express.
Database: MongoDB.
Voice Infrastructure: Twilio and Ultravox.
Authentication: JWT-based role isolation, HttpOnly cookies, and token revocation.
AI Agent: Hindi-first multilingual voice agent with secure healthcare coordination tools.

We implemented transactional appointment booking, patient ownership enforcement, two-phase confirmation for sensitive actions, outbound reminders, inbound voice access, missed-call callbacks, and clinician supervision dashboards.

Challenges We Faced

One of our biggest challenges was creating a reliable voice-first experience that works without requiring patients to navigate applications or keypad menus.

We addressed this through agent-first voice conversations, bilingual prompting, secure webhook verification, atomic callback reservations, bounded retries, and fail-closed authentication.

Testing real telephone calls also presented challenges due to Twilio trial restrictions and telephony credit requirements. We developed deterministic mock testing to verify workflows safely without unnecessary telephone charges.

What We Learned

We learned how to integrate conversational AI with real-world healthcare workflows, design secure patient authentication, implement reliable appointment transactions, and build telephony systems that remain safe during failures.

CareLoop represents our vision of healthcare coordination that is conversational, accessible, and inclusive.