Our problem statement is AI Disaster Assistant.

During disasters such as floods, earthquakes, fires, and extreme weather, one of the biggest problems is the lack of timely, accurate, and location-specific information. People may not know whether they are in danger, where the nearest safe place is, or what action they should take. At the same time, emergency teams have to process huge amounts of information from different sources.

Our solution is an AI-powered disaster assistant that can bring this information together and turn it into useful, actionable guidance.

The system can analyze disaster-related data such as weather conditions, alerts, satellite or map information, and user reports. Based on this data, it can identify high-risk areas, provide early warnings, answer disaster-related queries, and support emergency response.

Now, one of our key challenges is Best Use of Gemma 4.

Instead of depending completely on large cloud-based AI models, we want to make effective use of a lightweight open model. The idea is to use Gemma for tasks where fast and efficient AI processing is important—for example, understanding emergency messages, summarizing complex disaster information, classifying reports, and generating clear instructions for users.

This is especially useful during disasters because internet connectivity may be limited or unreliable. A lightweight model can help us build a system that is more efficient and potentially capable of operating closer to the user or on resource-constrained infrastructure.

So, our goal is not just to build another chatbot. We want to build an intelligent disaster-response assistant that can convert complex and scattered disaster data into clear, timely, and location-aware decisions that can help people respond faster and more safely.