SafeCompass

• 

• 

• 

Inspiration

Natural disasters don’t just destroy infrastructure — they create confusion. In the first minutes of a crisis, people often don’t know where to go, who to call, or which information is reliable. Even when help exists, it is scattered across websites, agencies, and languages. We were inspired by the idea that the biggest problem during disasters isn’t lack of resources — it’s lack of navigation.

What it does

CrisisCompass is an AI-powered disaster resource navigator that guides people to the nearest safe help during emergencies.

Users describe their situation in simple language. The AI:

identifies the type of crisis

prioritizes the safest response

routes users to nearby shelters, food, and medical aid

translates instructions into local languages

generates step-by-step survival guidance

highlights urgent risks

adapts for low-literacy users

Instead of searching across multiple systems, users receive one clear action plan.

Think of it as GPS for survival.

How we built it

We designed CrisisCompass as a modular AI decision engine:

NLP pipeline for crisis classification

routing logic to simulate safe evacuation paths

multilingual translation layer

resource prioritization model

explainable AI output for transparency

lightweight UI optimized for low connectivity

We simulated disaster datasets, emergency resource maps, and user scenarios to test decision accuracy. The architecture emphasizes speed, clarity, and fairness — prioritizing vulnerable populations in high-risk zones

Challenges we ran into

The biggest challenge was designing AI that remains responsible under pressure. In crisis scenarios, vague or incorrect guidance can cause harm.

We had to:

balance speed vs accuracy

design ethical decision logic

prevent hallucinated instructions

simplify complex emergency data

simulate real-world constraints like weak internet

Building trust into AI was harder than building the AI itself.

Accomplishments that we're proud of

Created a working prototype that produces clear emergency action plans

Built a multilingual crisis interface

Designed an AI prioritization system for vulnerable users

Simulated real disaster routing scenarios

Focused on ethical and explainable AI decisions

Built something that could realistically save lives

Most importantly: we transformed AI from a convenience tool into a safety tool.

What we learned

We learned that AI for good is not about flashy features — it’s about responsibility.

We explored:

ethical AI design

human-centered interfaces

crisis communication psychology

fairness in algorithmic decisions

real-world deployment challenges

This project showed us how technology must adapt to human vulnerability, not the other way around.

What's next for SafeCompass

Next steps include:

integrating real emergency APIs

offline functionality for disaster zones

partnerships with NGOs and governments

accessibility improvements for low-literacy users

voice-first interaction

real-time disaster feeds

field testing with community organizations

Our goal is to evolve SafeCompass into a globally deployable crisis infrastructure layer.

Because in emergencies, direction saves lives.

Built With

• algorithms
• data
• flask
• json
• language
• learning
• machine
• models
• natural
• processing
• python
• routing
• simulation