GuidedHealth

Inspiration

Living with chronic conditions can often be confusing and overwhelming for patients, with complex medical jargon and unclear guidance on managing their health. I was inspired to build GuidedHealth as a privacy-first, AI-powered companion that simplifies healthcare journeys by providing clear, personalized, step-by-step care plans. My goal was to empower patients and caregivers with trustworthy, accessible information at their fingertips.

What it does

GuidedHealth generates personalized care journeys for chronic illnesses based on user input. Leveraging a locally-run AI model specialized in medical understanding (MedGemma via Gaia Node) combined with live FDA drug information, it delivers a friendly, easy-to-follow roadmap of diagnosis milestones, lifestyle advice, medication insights, and decision points. The app also displays up-to-date drug safety info and includes disclaimers to promote informed, professional consultation.

How I built it

I built GuidedHealth as a full-stack web application using an Express.js backend coupled with a Bootstrap-based responsive frontend. The AI inference runs fully locally via Gaia Node hosting the MedGemma model, ensuring privacy and low latency. Drug data is fetched on-demand from the FDA’s public API to enrich medication info in the care journeys. I focused heavily on privacy, usability, and transparency - showing used AI model info and maintaining an educational disclaimer. The app includes helpful default prompts for quick use.

Challenges I ran into

One major challenge was extracting and validating medication names from AI-generated text reliably to query the FDA API without producing irrelevant or confusing results. I addressed this by implementing a whitelist-based approach and improving filtering. Integrating two different APIs (local AI + external FDA) smoothly to present coherent, well-formatted user output also required careful backend design. Lastly, balancing informative detail with user-friendly readability in the care journey was a continual effort.

Accomplishments that I'm proud of

I successfully built a polished, privacy-first healthcare guide that runs locally, combining sophisticated AI-generated personalized care plans with authoritative, real-time FDA drug safety info. The UI is clean, responsive, and intuitive on mobile and desktop, highlighting model transparency and including easy-to-use example prompts. The app meets key hackathon judging criteria comprehensively, delivering a meaningful patient empowerment tool ready for demonstration and future expansion.

What I learned

For me, I deepened our understanding of specialized medical AI models and how to leverage AI responsibly in healthcare contexts. The importance of privacy and transparency in patient-facing apps was reinforced throughout development. I also learned intricacies of integrating public health APIs like the FDA's for real-time data, managing text extraction challenges, and how to craft user-friendly clinical narratives. Finally, I honed rapid UI/UX prototyping with Bootstrap to build accessible, clean interfaces.

What's next for GuidedHealth

My roadmap includes multi-language support to reach diverse populations and printable/exportable care journey reports for offline use. I also plan to integrate with local personal health record stores securely to expand context-aware guidance. Additional condition-specific templates and symptom triage modules will enhance personalization. I'm also exploring offline AI inference capabilities to make GuidedHealth usable without internet connectivity - further strengthening privacy and accessibility for all users.

Built With

• express.js
• fda
• gaia
• node.js