Alert AI

Faster Response. Smart Guidance. Safer Lives.

Comment

Inspiration

The project idea was inspired by a personal childhood experience. At age eight, I suffered recurring seizures from a previous head injury. One day at school, I had a seizure, and the teachers tried to help but actually put me in more danger for example, placing a metal spoon in my mouth. Fortunately, a passerby who was a medical professional intervened correctly, provided proper first aid, and rushed me to the hospital. That experience made me realize how many people lack knowledge to respond safely during emergencies. When this hackathon opportunity came up, I knew I could build a solution to bridge that knowledge gap and improve emergency response.

What it does

Alert AI detects emergencies in real-time, verifies them using Gemini 3, notifies nearby users, and provides step-by-step guidance to ensure their safety until professional responders arrive. It also automatically alerts the appropriate emergency service, such as fire, medical, or police, based on the type of incident. Alerts are location-aware so only relevant users receive notifications, minimizing unnecessary panic.

How I built it

  • Edge device: Runs multiple emergency detection models (fire, bleeding, car accidents, etc.) connected to CCTV cameras.
  • Server: Flask backend handles verification, routing, logging, and proximity-based notifications.
  • Gemini 3 integration: Performs emergency verification and generates adaptive guidance instructions.
  • App: Receives alerts and guides users in real-time.
  • Database: SQLite logs emergencies and nearby user information for testing.

Example distance calculation using Haversine formula:

# Haversine formula
dlat = lat2 - lat1
dlon = lon2 - lon1
a = math.sin(dlat/2)**2 + math.cos(lat1) * math.cos(lat2) * math.sin(dlon/2)**2
c = 2 * math.asin(math.sqrt(a))

# Radius of earth in meters
r = 6371000
return c * r

Challenges I ran into

  • Integrating Gemini 3 verification without overloading the system.
  • Designing GPS-based proximity alerts to notify only relevant users.
  • Ensuring the demo flow is simple, clear, and reliable under tight deadlines.

Accomplishments that I’m proud of

  • End-to-end working demo: detection, verification, alerting, and guidance.
  • Successfully integrated Gemini 3 for verification and user guidance.
  • Built a scalable architecture ready for multiple emergency types.
  • Clear and professional system demonstration despite hardware constraints.

What I learned

  • How to orchestrate a multi-agent system combining edge devices, server, and app layers.
  • Strategies for designing proximity-based alerting and real-time guidance.
  • The importance of clear demo flow and concise storytelling for judges.
  • How to leverage Gemini 3 for practical, real-world AI applications.

What's next for Alert AI

  • Expand support to additional emergency types.
  • Deploy on live edge devices for real-world detection.
  • Integrate Firebase Cloud Messaging for production-grade alerts.
  • Continuously improve Gemini-guided instructions for more nuanced and adaptive guidance.

Built With

Share this project:

Updates