Perplexacare

Inspiration

We’ve all experienced the panic of Googling symptoms and being misled by unreliable or extreme diagnoses. I once searched for a minor issue and ended up with a terrifying result that wasn’t even close to accurate. That moment inspired me to build Perplexacare- a smarter, safer way to understand your health.

What it does

Perplexacare is an AI-powered health assistant that provides deeply personalized, real-time answers to your health and medication-related queries, tailored just for you.

Unlike other AI tools, Perplexacare:

Understands your medical history and user profile
Fetches verified medical data in real-time using advanced LLMs
Delivers relevant, context-aware responses

Features

User Profiles: Users set personal health details like age, allergies, and conditions to enable contextual answers.
AI-Powered Agent: Built using OpenAI’s Agent SDK, with Perplexity’s Sonar APIs (Sonar Reasoning & Sonar Pro) for real-time, source-backed information from the CDC, the Mayo Clinic, and other trusted medical sources.
Secure & Private: Uses Firebase for secure authentication and storing user profiles.

How I built it

I built Perplexacare using Next.js for the frontend to ensure a fast, responsive user experience and a lightweight Node.js backend to handle API communication. The core AI agent was developed using OpenAI’s Agent SDK, integrated with Perplexity’s Sonar Reasoning and Sonar Pro APIs to fetch accurate, real-time medical information from trusted sources like CDC and Mayo Clinic. For user authentication and storing personalized health profiles, I used Firebase, enabling secure, contextual conversations tailored to each individual.

Challenges I ran into

One of the main challenges I faced was deploying the frontend on Vercel, where API calls taking longer than 30 seconds would time out, especially when querying the LLM. I resolved this by optimizing the request structure and offloading heavy processing to the backend. Another key challenge was crafting robust and efficient prompts to ensure the AI consistently delivered accurate, personalized, and medically relevant responses

Accomplishments that I'm proud of

I was able to finally deploy the working app.

What I learned

Through building Perplexacare, I learned how to effectively integrate LLMs with real-world applications, especially in sensitive domains like healthcare. I gained hands-on experience with OpenAI’s Agent SDK, Perplexity Sonar APIs, and managing user-specific context in a secure and scalable way using Firebase. Additionally, I deepened my understanding of prompt engineering, API optimization, and the challenges of deploying AI-powered apps in production environments like Vercel.

What's next for Perplexacare

Deeper Contextual Memory: Enhance the chat interface to retain a richer understanding of past conversations for more accurate, ongoing support.
Doctor Transcription Integration: Allow users to scan and store doctor-provided prescriptions or transcriptions for the AI to reference during chats.
Medical History Awareness: Factor in the user’s complete medical history to offer safer and more personalized responses.
Scan Storage & Analysis: Provide a secure space for users to upload medical scans (e.g., X-rays, lab results), which the AI can analyze for better health insights.