Guardian Tweet

During a natural disaster, social media often becomes the fastest source of real-time information. For State Farm's 'Good Neighbor' challenge at Southern Hacks 2020, my team and I tackled this reality: how could we harness the speed of social media to create a proactive safety alert system?

Our 1st-place winning app, Guardian Tweet, did just that. I helped develop an NLP pipeline that ingested a live feed of tweets, using a Naive Bayes model we trained on disaster-specific language to automatically flag credible threats in real-time.

The system's true innovation was its user-centric alert process. When a disaster was detected, the app used the tweet's geolocation to alert users in the area. It then initiated a three-step safety check-in: first asking if the user was safe, then notifying their loved ones, and finally, providing tailored emergency resources to those who needed help. This transformed raw social media data into a powerful tool for community safety.