CrisisLens: Building a Personal Emergency Intelligence Radar for My Friend Arjun

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

🔍 CrisisLens: Personal Emergency Intelligence Platform

Transforming chaotic, conflicting disaster information into clear, verified, and personalized action.

---

What I Built

Who I Built It For

My friend Arjun is a university student living in Mohali (Sector 70, Punjab, India) who commutes daily along Airport Road to Chandigarh University.

The Problem

During heavy monsoon storms, Arjun faces extreme information overload:

• Panic in group chats: Viral WhatsApp forwards claim dams have collapsed.
• Conflicting travel alerts: Neighbors report Airport Road is open, while police issue an emergency closure notice 15 minutes later.
• Vague weather bulletins: Regional forecasts lack hyper-local street information.

Arjun is left with three urgent questions: Is my route flooded? Which claims are verified? What should I do right now?

The Solution: CrisisLens

CrisisLens synthesizes scattered news, official advisories, and community dispatches into a single, trustworthy intelligence brief tailored to Arjun's saved locations:

• Four-Tier Trust Model: Classifies every claim as CONFIRMED (green), REPORTED (amber), UNCERTAIN (gray), or CONFLICTING (red).
• Contradiction Detection: Flags conflicting reports and uses timestamps to determine which alert takes precedence.
• Personalized Impact: Evaluates hazards against Arjun's Home (Sector 70), College, and Commute, suggesting safe reroutes (e.g., Kharar bypass).
• Hands-Free Audio: Broadcasts life-critical summaries via voice for drivers or low-visibility situations.
• Interactive Map: Displays flood choke points, hazard radii, and safe detours on an interactive Leaflet map.
• Multi-Scenario Engine: Ready-to-demo scenarios for Mohali Floods (Arjun), Delhi Toxic Smog (Priya), and Himalayan Landslides (Vikram).

---

Demo

• Quick Local Demo: The project includes a zero-dependency demo engine that runs out of the box with simulated multi-disaster scenarios (no API keys required):

bash
  git clone https://github.com/divyamchoudhary1604/crisislens.git
  cd crisislens
  npm run dev

## Code


  

    

      
      
        divyamchoudhary1604
       / 
        crisislens
      
    
    

      CrisisLens - Personal Emergency Information Intelligence Platform. Built for Hacktoberfest 2026.
    
  
  

    

🔍 CrisisLens



“From scattered crisis information to clear, trustworthy context and action.”
Built for the Hacktoberfest 2026 — DEV “Build for a Friend” Challenge.













🎯 1. The Story: Built for a Friend



Meet Arjun


Arjun is a college student living in Mohali Sector 70 (Punjab, India) who commutes daily to Chandigarh University via Airport Road.


The Real Problem



During heavy monsoon storms and flash flooding, Arjun faces a flood of conflicting, unstructured data:



WhatsApp groups circulate frightening forwards claiming dams have collapsed.

Local news portals publish sensational headlines with outdated timestamps.

IMD weather bulletins issue broad regional warnings lacking localized guidance.

Neighbors claim Airport Road is open; meanwhile, district police quietly tweet an emergency closure.


Arjun is left paralyzed with three critical questions:




“Does this storm directly affect my home or college route right now?”

“Which claims are officially verified versus unconfirmed rumors?”



“What specific action should I take…




  


  
View on GitHub






Frontend: React 19, Vite, Leaflet, Responsive Dark Glassmorphism CSS

Backend: Node.js, Express, REST APIs

Database: MongoDB Atlas (with automatic in-memory fallback)

AI: Google Gemma 2 (google/gemma-2-9b-it) via OpenRouter, Groq, or local Ollama

Audio & Search: ElevenLabs TTS and SerpApi Google News




  
  
  How I Built It


CrisisLens uses an open-weight AI pipeline powered by Google Gemma 2:



Extraction: Parses raw bulletins and community posts into structured JSON entities and coordinates.

Conflict Resolution: Cross-examines sources, identifying contradictions and prioritizing newer authoritative directives over older rumors.

Personalization: Cross-references active hazard zones with user-saved waypoints to generate specific safety actions.

Flexible Deployment:



Hosted: Low-latency inference via Groq or OpenRouter.

Local / Edge: 100% offline inference with Ollama (gemma2:9b), ensuring the app works even when disaster strikes local internet infrastructure.

Demo Mode: Built-in zero-dependency fallback so anyone can clone and test instantly without API keys.






  
  
  Why Does Open Innovation Matter?


In emergency response, open innovation is essential:



Offline Survival: When natural disasters sever undersea cables and cell towers, closed-API models fail. Open-weight models like Gemma run locally on laptops, rescue vehicles, and emergency field servers.

Zero-Hallucination Auditability: Closed models change without notice. In life-or-death situations, prompts, outputs, and trust classifications must be transparent, verifiable, and auditable.

Equitable Access: Community responders and students cannot be locked out by subscription paywalls, API rate limits, or sudden platform bans during a crisis.




  
  
  🤖 AI-Assisted Development


CrisisLens was built collaboratively with an AI pair-programming agent. AI assistance was used throughout the development process for architecture, implementation, debugging, API integration, UI improvements, and documentation.

The final application integrates the Gemma-powered synthesis pipeline, Leaflet GIS mapping, MongoDB Atlas persistence, Express APIs, and Render deployment configuration.

  
  
  Prize Categories




🌟 Best Use of Gemma ($200): Gemma 2 (google/gemma-2-9b-it) powers the 10-step synthesis pipeline, 4-tier confidence classification, contradiction detection, and grounded safety recommendations.

🌟 Best Use of Render ($200): Full Infrastructure-as-Code blueprint (render.yaml) deploying the Express backend and React 19 static client in a single click.

🌟 Best Use of MongoDB Atlas ($100): Cloud persistence for multi-source alerts, citizen hazard reports (POST /api/reports), audit logs, and user location profiles.

🌟 Best Use of ElevenLabs ($100): High-clarity emergency voice synthesis (eleven_multilingual_v2) for hands-free audio briefings.

🌟 Best Use of SerpApi ($100): Real-time Google News search integration to ingest breaking advisories dynamically.




  
  
  💬 Let’s Discuss!



How does your local community handle emergency communications during floods, storms, or severe weather?
Have you experimented with running open-weight models like Gemma 2 locally with Ollama for mission-critical or offline applications?


Drop your thoughts and questions in the comments below!