UnityAid

Homepage Dashboard

Inspiration

UnityAid was inspired by the chaos of disaster response, where reports are often vague and incomplete. We wanted to create a system that could bring clarity quickly and help responders prioritize what truly matters.

What it does

UnityAid takes freeform reports, uses AI to classify their urgency, and asks clarifying questions when confidence is low. It then turns those inputs into structured tickets with mapped locations and live dashboards.

How we built it

We built UnityAid with Streamlit for the frontend, Folium maps for geolocation, and a custom AI triage agent powered by Google ADK + Gemini. The agent applies heuristics and conversational logic:

$$ \text{If } P(\text{classification}) < 0.7 \;\; \Rightarrow \;\; \text{Ask Clarifying Questions} $$

This blend of AI and real-time visualization allowed us to move from raw text to actionable insights.

Challenges we ran into

Some key challenges included session state persistence in Streamlit, making Folium maps responsive, and designing AI questions that improved accuracy without overwhelming the user.

Accomplishments that we're proud of

We created a system that can take a vague report like “someone needs help” and refine it into a priority 5/5 emergency with over 90% confidence after a few clarifying questions.

What we learned

We learned how to integrate conversational AI into real-time systems, balance AI confidence with human oversight, and design heuristics that meaningfully adjust priority classifications.

What's next for UnityAid

Next, we want to add persistent storage, integrate Redis pub/sub for multi-worker scaling, and test UnityAid against more real-world scenarios to validate its effectiveness in high-stress environments.