POSEIDON

• 

  Poseidon

Inspiration

Poseidon AI Stay safe from floods. Get early warnings, understand your risks, and know when to evacuate.

What it does

Early Flood Warnings Get alerts about incoming typhoons, heavy rains, and flood risks in your area. Know when to prepare, when to stay alert, and when to evacuate your family to safety.

How we built it

The core of Poseidon is built on a modern, scalable tech stack:

Google Gemini API: Used for natural language processing to synthesize complex weather reports into simple, calming instructions and to power our safety chatbot.

Google Maps Platform: Utilized for geospatial visualization and calculating evacuation routes that avoid known flood-prone lowlands.

Firebase: Provides the backend infrastructure, including Firestore for real-time data synchronization and Firebase Cloud Messaging (FCM) for low-latency emergency push notifications.

OpenWeather/Public Agency APIs: We integrated global and local meteorological data streams to feed our risk-assessment model.

Challenges we ran into

Data Latency: In a flood, every second counts. We initially struggled with the delay in processing large weather datasets. We solved this by implementing a "priority-queue" system in Firebase that pushes critical alerts before processing secondary analytics.

UX for Crisis: Designing an interface for someone who might be panicked or in low-light conditions was difficult. We had to strip back our original design to focus on high-contrast, large-scale UI elements and "One-Tap" safety check-ins.

Topographical Accuracy: Standard maps don't always show micro-elevations. We had to research and integrate specialized elevation APIs to ensure our "safe zones" weren't actually in basins.

Accomplishments that we're proud of

Seamless AI Integration: We successfully tuned the Gemini model to provide "Crisis-Mode" responses—concise, authoritative, and empathetic—avoiding the long-windedness typical of LLMs.

Real-time Synchronization: Achieving sub-second alert delivery across multiple test devices.

Accessibility: Building a system that remains functional and legible even on older mobile devices with limited processing power.

What we learned

The Power of Simplicity: In emergency tech, "less is more." We learned to prioritize the most critical data points over "cool" but unnecessary features.

Geospatial Logic: We gained a deep understanding of how to manipulate GPS data and elevation gradients to predict water flow patterns.

Community Trust: We learned that for an AI to be effective in a disaster, the user needs to trust its logic, which led us to include "Why this alert was sent" transparency features.

What's next for POSEIDON

Predictive Hydrology: Moving from "reactive" alerts to "predictive" modeling using historical flood data and machine learning to forecast street-level flooding 12 hours in advance.

Government Integration: Partnering with local disaster management offices to pipe Poseidon data directly into official emergency response dashboards.

Built With

• gemini