Back to Projects
FinishedHackathonCase study

MedAlert AI

A full-stack hackathon submission that aggregates crisis signals from earthquake, weather, news, and social sources, lets analysts verify findings in a live dashboard, and pushes AI-drafted alerts in five languages to affected users.

Problem

During a developing crisis, raw signals — seismic feeds, weather APIs, news, social posts — arrive faster than anyone can verify them, and affected people rarely get timely alerts in their own language.

Solution

The backend aggregates feeds from USGS, EMSC, weather and news APIs, and social sources into intel findings that stream to an analyst dashboard over Socket.IO. Analysts can escalate a finding into an alert, run deeper multi-source verification through Perplexity, and publish — with Gemini generating alert copy in Arabic, English, Turkish, Italian, and Hebrew, delivered as browser push notifications with Haversine-based geo targeting.

What I learned

In a hackathon, resilience beats completeness — building demo-data fallbacks for every external service meant the product stayed demoable even when APIs rate-limited us mid-presentation.

Tech Stack

ReactViteExpressSocket.IOSequelizeSQLiteGeminiPerplexity

Screenshots

More Hackathon projects

Hackathon

Hackathon

Hackathon Campus Map

A quick hackathon prototype for finding rooms, booths, and event checkpoints.