Apexvision - real-time, on-device AI navigation for blind & low-vision users

ApexVision is a smartphone app that gives people who are blind or have low vision real-time guidance about the world in front of them - identifying obstacles, describing surroundings, and helping them move safely through a space, with results spoken in under 200 milliseconds.

Why it matters: 2.2 billion people worldwide live with vision impairment. Existing tools force a hard tradeoff - dedicated hardware costs $3,500–$4,500, popular apps stream a user's live camera to remote strangers, and cloud-based tools add seconds of delay that make them unsafe at a street crossing. Roughly half of users abandon assistive devices altogether because visible hardware publicly marks them as disabled. ApexVision removes that tradeoff.

How it works: Four computer-vision models - semantic segmentation, depth estimation, object detection, and a vision-language model - run entirely on the user's existing iPhone, with no internet connection and no data ever leaving the device. This makes the experience real-time, fully private, and functional offline. Three modes serve different needs: Finder locates a specific object on request, Safety highlights walkable paths and flags hazards, and Descriptive narrates the surrounding scene on demand.

Proof to date: ApexVision is live on iOS. The underlying technology originated in University of Tennessee computer-vision research, is published in IEEE CVPR, and won the NVIDIA AI City Challenge. It is protected by U.S. Provisional Patent 63/917,447. We completed an institutional pilot with Signal Centers, an assistive-technology center in Chattanooga, which confirmed the real-time navigation capability was new to their program and valued by users. ApexAI was selected for Cohort 1 of the NSF-funded NobleReach Science to Venture program, and completed NSF I-Corps customer discovery.

Next steps: We are expanding to newer LiDAR-equipped devices and to Android, growing institutional pilots with state vocational rehabilitation agencies and assistive-technology centers, and developing a predictive safety layer that anticipates hazards and conveys its own confidence. We are raising a $1M seed round and seeking partners in accessibility distribution, institutional procurement, and edge-AI hardware.