TeleHealth AI

Inspiration

The idea for TeleHealth AI was inspired by the growing need for accessible, personalized healthcare guidance, especially in remote areas where specialist access is limited. We wanted to leverage recent advances in AI to provide trustworthy, evidence-based medical advice in a scalable way.

What I Learned

During the development of this project, I gained deep experience in integrating multiple AI services like Perplexity’s Sonar API and Google Gemini, as well as building robust backend systems with Prisma and tRPC. I also enhanced my skills in Next.js 14 with the App Router, TypeScript, and building responsive, accessible UI using Tailwind CSS and shadcn/ui.

How I Built It

The platform was built using a modern full-stack approach:

Frontend developed with Next.js 14 and React 18, focusing on performance and user experience.
Backend APIs implemented via tRPC, connected to a PostgreSQL database managed with Prisma ORM.
AI integration with Perplexity Sonar API to deliver citation-backed medical responses.
Authentication handled securely with Clerk.
Caching implemented through Upstash Redis for improved performance.
Deployed on Vercel for seamless continuous deployment.

Challenges Faced

Dependency conflicts in the Next.js ecosystem required careful resolution to maintain compatibility across libraries.
Ensuring medical information accuracy and compliance with privacy considerations involved integrating citation APIs and implementing robust data filtering.
Designing a multi-persona AI system demanded thoughtful orchestration of conversational flows to provide relevant specialist advice.
Balancing real-time voice input with text-based chat while maintaining accessibility and usability.