VaiSight
AI Accessibility Vision
Overview
I built VaiSight as an AI accessibility prototype for camera-enabled glasses. The idea was to use computer vision, OCR, object recognition, scene reading, audio output, and API-backed processing to help blind and visually impaired users understand the world around them.
This project matters to me because it shows AI being used for a practical human need. The goal was not to build a flashy demo. The goal was to explore how software could help someone read prices, identify objects, understand surroundings, and receive useful spoken guidance.
What I built
I designed the prototype around a simple loop: capture visual context, process it through AI and computer vision tools, translate the result into useful language, and deliver it through audio. That required thinking about speed, clarity, error handling, and the difference between technically correct output and actually helpful guidance.
The system combined OCR for text, scene understanding for environmental context, object recognition for everyday identification, and voice output so the experience could work hands-free.
Accessibility as product design
Assistive technology has to be careful. The user is not asking for a novelty response; they need information they can act on. I approached VaiSight with that in mind, focusing on practical use cases like reading labels, identifying household items, understanding a room, and supporting voice-guided movement.
VaiSight is one of the clearest examples of how I think about AI: the model is only one part of the system. The product design, interface, latency, output format, and safety assumptions matter just as much.