Pulse - AI assisted ground control

Alert testing
Design Block Diagram and Overview

Inspiration

Large festivals can have thousands of attendees and hundreds of volunteers spread across different zones. During stressful incidents, reports can be incomplete, unclear, or difficult to coordinate for a limited amount of managers. We wanted to help safety leads turn messy frontline reports into clear, actionable information, and make the jobs of the managers easier.

What it does

Pulse is an AI-assisted incident coordination platform for festivals. Volunteers can submit voice or text reports, while Pulse's LLM extracts key details, identifies missing information, asks follow-up questions, and recommends suitable responders. Managers can then review these key details, make necessary changes and send out alerts to the qualified volunteers nearby to resolve the situation.

Safety leads can review incidents, approve or modify recommendations, track updates, send alerts, and resolve incidents. The final decision always remains with a human.

How we built it

We built Pulse using:

React, TypeScript, and Vite for the web dashboard
Expo and React Native for the volunteer mobile app
FastAPI and Python for the backend
SQLite for incident and resource data
Local Whisper speech recognition for voice reports
The system uses deterministic responder scoring based on qualifications, availability, location, and current assignments. This makes recommendations predictable and explainable for the prototype.

Challenges we ran into

One challenge was designing AI assistance without allowing it to make safety-critical decisions. We made sure the system clearly distinguishes between a recommendation and an approved response.

We also had to handle incomplete reports, simultaneous incidents, responder conflicts, and different needs between volunteers and safety leads.

Accomplishments that we're proud of

We built an end-to-end workflow from:

Volunteer report → AI understanding → clarification → recommendation → human approval → response tracking → resolution

We are particularly proud of the separate mobile volunteer experience, manager dashboard, explainable responder recommendations, incident timeline, voice reporting, reviewed alerts, and simulated festival map.

What we learned

We learned that AI is most useful when it reduces information overload rather than simply acting as a chatbot. Structured extraction, clarification, summarisation, and recommendation explanations can help safety teams make faster decisions.

We also learned that predictable rules are important for resource allocation, while AI is well suited to understanding natural language and explaining complex situations.

What's next for Pulse - AI assisted ground control

Next, we would add real-time updates, stronger duplicate and escalation detection, configurable festival procedures, offline mobile support, real GPS tracking with consent, and more advanced incident pattern detection.

Our goal is to make Pulse an AI-assisted ground control platform that helps event teams understand incidents, coordinate resources, and stay in control when pressure is high.