bystander

An emergency situation presents itself. Take an 8-second clip, and get immediate and actionable help at your specific location.

Comment

Inspiration

You are a bystander. An emergency situation is going down, directly in front of you. But there's nothing you can do, because you don't know what to do. You feel powerless.

In one year, over 9 million Canadians were bystanders in an emergency situation, and weren’t trained in first aid. 36% of Canadians say that no one in their household is first-aid trained

Consider a person being deprived of oxygen. Within 4 to 6 minutes, their brain begins to suffer irreversible damage.

Immediate action is crucial because these are minutes where panic, confusion, and poor communication can make everything worse.

Bystander exists for that moment. Not to replace professional help. But to bridge the gap until they arrive.

Startup Potential

We pitched this project to the security guard in SPUR Innovation Centre. He loved it, and said he would use it in his job.

In doing this, we showed that the product is not limited to healthcare. There is potential for real world adoption in different industries. It has applications in crime and security.

What It Does

  1. Capture — Record or upload an 8 second clip of the emergency scene to Cloudinary
  2. AI Scoring — Gemini 2.5 Flash analyzes the scene to assess injuries, consciousness, bleeding, and body position, producing a severity score from 1-10
  3. Biometric Vitals — Presage SmartSpectra extracts heart rate and breathing rate directly from the video, no wearables needed
  4. Dashboard — Displays a full situation breakdown: severity banner, detected injuries, risk flags, and prioritized first aid steps
  5. AI First Aid Coach — Chat or use voice mode to get real-time, directive guidance from an AI powered by Gemini and ElevenLabs
  6. Crowdsource Alerts — When severity hits 8+, registered first aiders in the area are automatically notified via Twillio SMS with incident details and location
  7. Incident Report — Generates a printable ER-ready report with all data and chat transcript

How It Works

Camera/Upload → Frame Extraction → Gemini Vision Analysis
                                  ↘
                Video → Cloudinary → Presage SmartSpectra + Gemini (Docker env)
                                  ↘
                        Merged Triage Data → Dashboard
                                           → SMS Alert (if critical)
                                           → AI Coach (Gemini + ElevenLabs TTS)
                                           → Incident Report

How We Built It

  • Frontend: React 19, Vite, Tailwind CSS 4, Framer Motion
  • Backend: Node.js, Express
  • AI/ML: Google Gemini 2.5 Flash (vision + chat), Presage SmartSpectra (contactless vitals)
  • Voice: ElevenLabs (text-to-speech), Web Speech API (speech-to-text)
  • Media: Cloudinary (video upload, storage, playback)
  • Alerts: Twilio (SMS notifications)

Challenges

Getting Presage working was the defining challenge. Integrating a clinically validated C++ biometrics SDK into a JavaScript backend over a child process bridge required a full Docker build environment, switching from Spot mode to Continuous mode, and rebuilding the metrics callback architecture from scratch. One commit message sums it up: "finally fixed presage."

Running Gemini and Presage in parallel with a clean merge strategy also required careful orchestration to ensure output was consistent.

What's Next

  • Native mobile app for faster camera access
  • GPS tracking for precise location
  • 911 auto-dial with status packet pre-loaded

Built With

Share this project:

Updates