GuideLight

Inspiration:

Our inspiration came from realizing how inaccessible disaster relief is. In emergencies, information overload trap vulnerable people. For a blind individual, navigating dense legal text on a government aid form during a crisis is an immense, isolating barrier. We wanted to build a solution that feels like a calm friend guiding you through chaos.

What it does:

GuideLight acts as an intelligent voice bridge. It processes complex crisis papers, extracts critical milestones, and delivers a 3-tier spoken action checklist (What to bring, Where to go, Deadlines). It then cross-references the user's location to provide turn-by-turn audio navigation to nearby physical aid distribution centers.

How we built it:

We engineered an automated end-to-end cloud pipeline. The system inputs data via an optimized OCR layer, processes the text using a localized NLP framework coupled with specialized large language models, and outputs localized results using an audio-first recommendation layer integrated with mapping APIs.

Challenges we ran into:

The hardest problem was emotional design. For a blind user, a silent loading screen triggers immense anxiety. We bypassed this UX gap by designing an active audio loading state that provides continuous sound feedback while the backend engine is running. On the technical side, stopping model hallucinations under low-quality document scans required heavy filtering.

Accomplishments that we're proud of:

We are proud that GuideLight moves beyond a basic screen reader. It successfully takes a user from total confusion to physical, real-world relief in a single, stress-free voice interaction, keeping humans safely routed at every critical step.

What we learned:

We learned that accessibility is primarily an information problem before it is a physical one. Designing for the blind community requires treating system latency and interface feedback as safety features, ensuring the technology always speaks clearly and patiently.

What's next for GuideLight:

We plan to introduce multilingual integration for Arabic, Spanish, and French to directly support global refugee communities in coordination with the UNHCR. We are also building a specialized "Caregiver Mode" for families and caseworkers to co-manage crisis checklists.