Yelp4Help

Inspiration

While there are plenty of existing emergency alert systems built to handle widespread catastrophic events, everyone ignores what happens when local emergency rooms, police dispatchers, and first responders become entirely overwhelmed. Thousands of citizens with urgent, but not immediately fatal, needs get lost in the chaos while those in critical condition face lethal delays. We built Yelp4Help to fix this blind spot. By creating a decentralized routing system, we ensure that while high-risk patients are fast-tracked straight into the medical system, others are instantly matched with community mutual-aid supplies and actionable self-help, helping to avoid an overloaded emergency infrastructure.

What it does

Yelp4Help acts as an automated, community-driven crisis routing engine that steps in when 911 lines and emergency rooms get completely overwhelmed during disasters. By analyzing a user's crisis description in plain text, the platform dynamically acts as an air-traffic controller: it flashes immediate 911 instructions to high-risk users while automatically placing a live phone call to brief the nearest available ER on their exact trauma condition, instantly matches medium-risk users with localized maps of community mutual-aid supplies and open shelters so they aren't ignored in a backed-up queue, and provides low-risk users with step-by-step self-help survival coaching to safely resolve minor issues on their own—effectively offloading pressure from a collapsing emergency infrastructure when every second counts.

How we built it

Using Gemini and the AI App Builder on Google Cloud Platform, we created a model that semantically classifies situations into High, Medium, or Low risk. Medium-risk inputs trigger a localized resource lookup engine, while low-risk inputs trigger step-by-step solutions. High-risk immediately calls the nearest medical facility, notifying them of the patient's condition and urges the patient to call 911 immediately. We deployed this app through GCP.

Challenges we ran into

Our biggest hurdle was ensuring zero hallucination and reliability during high-stress scenarios since AI cannot be allowed to invent a fake hospital phone number or miscalculate a risk. We overcame this by providing real-time data regarding resources and centers.

What we learned

We learned that the true power of Large Language Models in production isn't just text generation—it’s semantic orchestration. Building this project taught us how to bridge the gap between abstract AI understanding and immediate, real-world physical infrastructure, proving that a cloud function can successfully turn an emergency text message into a ringing phone line that saves lives when seconds count.