Inspiration: During disasters, official alerts are often written at a college reading level, but 54% of US adults read below a 6th-grade level. We wanted to fix that. ClearSignal was born from the idea that life-saving information should be understandable by everyone, not just those who can parse bureaucratic language under stress.
What it does ClearSignal transforms emergency alerts into accessible formats in real time. Paste any alert text and get three simplified versions (Basic, Intermediate, and Comprehensive), translated into six languages (English, Spanish, French, Chinese, Arabic, Portuguese), with audio playback for each. A live crisis map pulls data from GDACS and NewsAPI to show geo-tagged emergencies worldwide as severity-coded markers, auto-refreshing every 5 minutes.
How we built it: React + Tailwind frontend with Mapbox GL JS for the interactive map Node.js/Express backend with TypeScript OpenAI for text simplification and translation in a single LLM pass Hume AI for natural-sounding text-to-speech GDACS RSS feed + NewsAPI for live crisis data Flesch-Kincaid readability scoring to validate each simplified variant Deployed on AWS Elastic Beanstalk (backend) and AWS Amplify (frontend) Challenges we ran into Getting the LLM to consistently produce three meaningfully distinct reading levels — not just slightly reworded versions of each other — required significant prompt engineering. Merging two async data sources (GDACS and NewsAPI) with graceful degradation while keeping the map markers accurate and the feed chronologically sorted proved trickier than expected. Accessibility was also a constant balancing act: ARIA live regions, keyboard navigation, and screen reader announcements had to work across every dynamic state change.
Accomplishments that we're proud of The readability pipeline actually works — Grade 3 variants consistently score below 4.0 FK, and Grade 9 variants below 10.0. The map and feed feel genuinely integrated: geo-tagged news articles show up as live markers. We also built a solid property-based test suite using fast-check that catches edge cases across hundreds of generated inputs, giving us real confidence in the correctness of the core simplification and feed logic.
What we learned LLMs are great at simplification but need tight output contracts — structured JSON responses with explicit level keys made parsing reliable. We also learned that accessibility isn't a feature you bolt on at the end; building ARIA support into the component design from the start saved us a lot of pain. And property-based testing is genuinely underused — it found bugs our unit tests missed.
What's next for ClearSignal Push notifications so users get alerts for their region without opening the app D.irect integration with FEMA's IPAWS and the Emergency Alert System Offline mode ,so the app works when cell networks are degraded during disasters Community-verified translations to improve accuracy for low-resource languages A public API so emergency management agencies can embed ClearSignal's simplification engine into their own systems
Built With
- aws-amplify
- aws-elastic-beanstalk
- express.js
- gdacs-rss
- hume-ai-(tts)
- mapbox-gl-js
- newsapi
- node.js
- openai-api
- react
- tailwind-css
- typescript
- vite
Log in or sign up for Devpost to join the conversation.