CrisisLink

Inspiration

In an emergency, people do not communicate in perfect sentences. They panic, speak in their native language, send incomplete messages, or may only be able to speak into their phone. Yet many emergency systems still depend on forms, manual location entry, language selection, and clear descriptions.

We built CrisisLink around one simple idea: during an emergency, technology should adapt to the person — not the other way around.

What it does

CrisisLink is an AI-powered multilingual emergency triage and volunteer dispatch platform that converts unstructured voice or text into a structured emergency response.

A victim can simply type or speak naturally in their own language. CrisisLink then:

• Understands multilingual voice and text using Google Gemini
• Translates and structures the emergency automatically
• Determines the severity and emergency category
• Extracts important symptoms and context
• Captures the victim's GPS location
• Reverse-geocodes coordinates into a readable address
• Determines the appropriate emergency number
• Provides one-tap emergency escalation
• Alerts nearby registered volunteer responders
• Sends responders structured information including severity, translation, and location

Instead of forcing someone in distress through multiple forms, CrisisLink turns a chaotic cry for help into actionable information within seconds.

How we built it

CrisisLink is built as a modern serverless web application using Next.js and React.

The user can submit an emergency through text or microphone input. The raw information is processed using Google Gemini's multimodal capabilities, which converts the input into structured emergency data.

For location intelligence, we use the browser's Geolocation API to obtain GPS coordinates and OpenStreetMap / Nominatim to reverse-geocode them into a human-readable location.

Once triage is complete, CrisisLink can generate the appropriate emergency escalation path and uses the Resend API to deliver structured alerts to volunteer responders.

The application is deployed live using Vercel.

Our simplified flow is:

Voice / Text → Gemini AI → Structured Emergency → GPS & Address → Emergency Routing → Volunteer Alert

Challenges we faced

One of our biggest challenges was handling emergency input that is naturally messy. People may mix languages, leave sentences unfinished, or provide information in a completely unstructured way.

We also had to make sure the AI returned predictable structured data instead of simply generating conversational responses.

Location handling introduced another challenge because raw GPS coordinates are not useful to most responders. We therefore integrated reverse geocoding to transform coordinates into usable street-level information.

Finally, connecting AI triage, geolocation, emergency routing, and responder notifications into one fast workflow required careful handling of asynchronous APIs and failure states.

What we learned

Building CrisisLink taught us that AI becomes far more useful when it is connected to real-world systems rather than being used only as a chatbot.

We learned how to combine:

• Multimodal AI
• Structured AI outputs
• Browser geolocation
• Reverse geocoding
• Serverless APIs
• Automated email dispatch
• Real-time emergency-oriented UI/UX

Most importantly, we learned to design around the user's situation. Someone facing an emergency should not need to understand the technology behind the system.

What's next

Our next goal is to evolve CrisisLink into a larger emergency-response network with:

• Distance-based volunteer matching
• Real-time responder availability
• Live responder tracking
• SMS and WhatsApp alerts
• Hospital and ambulance integration
• Offline / low-bandwidth emergency reporting
• More regional languages and dialects
• Emergency authority dashboards
• Automatic escalation when volunteers are unavailable

Our long-term vision is simple: make getting help as easy as asking for it.

Because in an emergency, the only language that should matter is help.

Built With

• ai
• api
• gemini
• google
• gps
• multimodal
• next.js
• nominatim
• openstreetmap
• react
• resend
• typescript
• vercel