LifeLine Buddy

Every Second Counts in Emergencies — LifeLine Buddy Responds when human agents can't!

Comment

🚨 LifeLine Buddy

For Demo : First Call the 911 simulated AI agent at 234-265-9301 and then visit the try out link given below to access the human agent side dashboard. (https://www.lifelinebuddy.us/)

🧠 Inspiration

911 dispatch centers in the U.S. handle over 240 million calls every year — but most are under-resourced, understaffed, and overwhelmed. In high-density cities like New York, Washington D.C., and San Francisco, callers can be left waiting for 50–60 seconds or more to reach a dispatcher — a dangerous delay in emergencies where every second counts.

This issue is worsened by the manual process of triaging calls. Human operators must quickly assess urgency from panicked or unclear speech, often without enough context or time.

We wanted to help solve this. LifeLine Buddy was built to ensure that no emergency call ever goes unanswered, and that dispatchers always have the context they need to make faster, smarter, and more equitable decisions.

🚀 What It Does

LifeLine Buddy is an AI-powered emergency dispatch assistant that steps in when all human 911 operators are busy.

  • It instantly answers incoming 911 calls using natural-sounding voice via
  • Gathers critical information like location, type of emergency, and caller emotional state
  • Analyze the urgency, emotion, and keywords in real-time
  • Classifies calls as Critical, Urgent, or Standard
  • Gives immediate, context-aware guidance
  • Displays all information — transcripts, summaries, severity, and emotion scores — on 911 agent dashboard
  • Suggests appropriate emergency response units (e.g., police, fire, EMS)

It doesn't replace dispatchers — it amplifies them.

🛠️ How We Built It

  • Frontend: Built using React, TypeScript, Tailwind CSS, and Framer Motion, designed to show:
    • Live call transcript
    • Emotion and urgency analysis
    • Suggested response units
    • Real-time call queue with prioritization
  • Voice + AI Conversation:
    • ElevenLabs API powers the AI voice that talks with callers
    • Gemini API handles both:
    • Conversational logic (gathering information)
    • Analyzing text for emotions, urgency, and emergency type

  • Backend:

    • Built with Python Flask
    • MongoDB Atlas stores call logs, severity scores, timestamps, and analytics
    • Socket.IO handles real-time communication between backend and dashboard
    • Twilio to host the AI agent number

  • Hosting:

    • Frontend with AWS Amplify
    • Backend with Heroku
    • Domain name from GoDaddy

We focused on building a modular, scalable system with real-time performance in mind — because emergency response can't afford lag.

🧗‍♀️ Challenges We Ran Into

  • Limited Time meant we couldn't train traditional models — so we leaned entirely on Gemini's API capabilities
  • Handling real-time API calls across ElevenLabs and Gemini, while ensuring low latency, was technically challenging
  • Balancing natural AI conversation flow with rapid data collection — without frustrating or confusing the user — took lots of iteration
  • Designing a UI that gave dispatchers all the context they need without overwhelming them
  • Building ethical safeguards to ensure AI suggestions never override human judgment

🏆 Accomplishments That We're Proud Of

  • Built a fully functioning deployed prototype that simulates real 911 call flows with AI-driven responses and prioritization
  • Achieved near-instant response time from AI assistant under simulated stress load
  • Designed a clean, usable dashboard that testers said felt intuitive and powerful
  • Successfully integrated voice-to-AI-to-dashboard pipeline end-to-end in under 48 hours
  • Created something that feels impactful, timely, and scalable to real-world use

📚 What We Learned

  • AI-powered conversation systems are finally fast and smart enough to assist in critical use cases — but they need thoughtful design and ethical safeguards
  • Emotion analysis from speech is not just a technical feature — it can be life-saving when paired with prioritization
  • Real-time systems require not only speed, but clarity — both for machines and humans
  • Designing for public safety means always putting the user and the victim first — simplicity, empathy, and speed matter more than fancy features

🔮 What's Next for LifeLine Buddy

  • More nuanced triage models trained on larger emergency datasets, including background noise and stress detection
  • Partnering with civic tech and public safety organizations to pilot the solution in real dispatch environments
  • Building out dashboard analytics for reporting trends, missed emergencies, and continuous feedback loops
  • Adding an offline fail-safe mode for areas with weak connectivity or sudden outages

We believe LifeLine Buddy can become a vital piece of the future of emergency response — ensuring that no call is left unanswered, and every second counts.

Share this project:

Updates