Inspiration
The idea for CareGuide was inspired by a common healthcare gap: patients often leave medical visits with limited understanding or recall of their care instructions. Meanwhile, rich clinical data sits within Electronic Health Records (EHRs), often underutilized. We envisioned a solution that could bridge this gap—by transforming static EHR data into actionable, spoken lifestyle guidance through a conversational AI interface. Our goal was to make personalized healthcare advice more continuous, accessible, and engaging.
What it does
CareGuide is a voice AI agent that delivers real-time, personalized lifestyle guidance using clinical data from EHRs. It interprets key health metrics—such as lab results, vitals, diagnoses, and care plans—and translates them into dynamic health tips, behavior nudges, and reminders. The AI agent speaks to users in natural language, offering context-aware advice on diet, activity, medication, and wellness goals, tailored to their medical history and current condition.
How we built it
We integrated several components to bring CareGuide to life: EHR data parsing using FHIR APIs to access structured patient data.
AI agent for interpreting clinical metrics and transforming them into layperson-friendly health advice.
Voice interface built using Groq and webRTC.
Decision engine powered by Rag AI agent to generate appropriate lifestyle suggestions based on clinical context.
The prototype was built using Python, Groq, Vector Data base , and cloud services for AI.
Challenges we ran into
EHR standardization: Variations in how providers structure and code data made consistent interpretation challenging.
Personalization vs. generalization: Striking the right balance between clinically accurate advice and user-friendly language was tough.
Voice UX tuning: Crafting natural, non-repetitive conversations with contextual intelligence took significant iteration.
Security & access: Compliance with HIPAA planned through obtaining patient consent through EHR sign in screen.
Accomplishments that we're proud of
Successfully created a working prototype that can read real clinical data and generate spoken lifestyle recommendations based patient id with real EHR server hosted by HAPI FHIR (hapi.fhir.org).
Built an intuitive voice interaction model that responds naturally to user questions and health contexts.
Developed a modular framework that can scale across different EHR systems and user demographics.
Maintained a privacy-first design without compromising user experience.
What we learned
Voice can be a powerful, empathetic interface when paired with the right data and context.
Patients are more likely to engage with health information when it’s delivered in simple, timely, and personalized ways.
The potential of EHR data goes beyond record-keeping—it can drive continuous care and behavior change with the right interface.
Collaboration across clinical, technical, and UX domains is essential for building effective healthcare tools.
What's next for CareGuide: Voice AI agent for EHR lifestyle tips
Our next steps include:
Pilot testing in real clinical settings with patient volunteers.
Expanding clinical logic to support more conditions (e.g., cardiac rehab, diabetes management, post-discharge care).
Enhancing multilingual support to broaden accessibility.
Integrating with wearable data for richer lifestyle insights.
Ultimately, we aim to turn CareGuide into a trusted voice companion for daily health—empowering people to live better, guided by their own clinical data.
Log in or sign up for Devpost to join the conversation.