CareLink

Inspiration

What if AI could handle the coordination, so healthcare workers could focus on care?

CareLink was born from that question — a platform that uses voice AI, computer vision, and generative video to bridge the gap between rural patients and their care teams.

What It Does

CareLink is a full-stack healthcare coordination platform with two portals:

For Clinic Staff

Voice-to-Referral — nurses speak naturally and AI auto-fills the referral form
Cloned Nurse Voice Calls — record once, generate unlimited personalized patient calls via ElevenLabs
AI Document Analysis — GPT-4o extracts diagnoses, medications, and lab values from uploaded docs
Real-Time Work Queues — live dashboard for appointment, callback, and ride requests

For Patients:

AI Weather Video Preview — upload a photo, Veo 3.1 Fast generates a video showing your environment
Interactive Weather Radar — Leaflet map with precipitation overlay and storm warnings
Referral Progress Tracking — 6-step visual pipeline
Voice Message Notifications — hear your nurse’s cloned voice delivering updates

How We Built It

Frontend:

Two React 19 + Vite apps
Leaflet maps
Web Speech API
MediaRecorder

Backend

Express.js
OpenAI GPT-4o
ElevenLabs voice cloning
Google Veo 3.1 Fast (@google/genai SDK)
OpenWeatherMap API

Challenges We Ran Into

Veo API integration was brutal — inlineData and fileData both rejected despite docs. Only the official SDK with imageBytes worked.
Voice cloning format compatibility between browser MediaRecorder and ElevenLabs
Real-time keyword matching for mapping spoken medical terms to form fields
OpenWeather free tier lacks alerts, so we built custom alert logic

Accomplishments That We’re Proud Of

Voice-to-referral pipeline: 15 seconds of speech replaces 3 minutes of typing
Cloned voice calls indistinguishable from real nurse calls
Weather video preview — seeing your own driveway in a simulated blizzard is far more impactful than “heavy snow expected”

What We Learned

AI APIs are moving targets — always use official SDKs over raw REST
Voice UX requires speech recognition, permissions, format compatibility, and feedback all working together
Rural healthcare has unique constraints that urban health tech ignores