Blind Assistive Navigation Software

My goal is to build a reliable software layer that lets a blind or visually-impaired user move confidently in both indoor and outdoor settings. Meta glasses and similar wearables already give us a camera and basic audio, yet they fail when background noise masks instructions and when no haptic feedback is available. I need a programmer who can rethink that whole flow and write the logic that ties the sensors, recognition, guidance, and feedback channels together so the experience is truly usable.

Core flow
• Real-time positioning: use camera, GPS, IMU and, where available, Wi-Fi/BLE beacons to pinpoint the user’s location and heading, whether in a shopping mall or on a city street.
• Path planning & turn-by-turn guidance: generate concise, context-aware instructions that avoid cognitive overload.
• Multimodal feedback: combine short audio prompts, optional bone-conduction output, and vibration/haptic cues to overcome noisy environments.
• Robust obstacle awareness: fetch depth or stereo data (or infer depth with AI) so the user is warned about hazards at cane-length distance or overhead.
• Extensible add-ons: hooks for voice commands, emergency SOS, text reading, object recognition, or any future computer-vision module.

I already have access to sample hardware (camera-equipped glasses and a tactile band), so you can prototype quickly with OpenCV, TensorFlow Lite, PyTorch Mobile or a stack you prefer. What I really need is the architecture, clean code, and demonstrable logic that meld everything into a smooth UX.

Deliverables
1. Source code with clear documentation and build/run instructions
2. A runnable demo (APK, executable, or Web build) that proves indoor & outdoor navigation on my test routes
3. API or module descriptions so more features can be plugged in later
4. A short technical report explaining algorithms chosen, sensor fusion method, and latency/accuracy benchmarks

Acceptance criteria
• Navigation prompts stay under 1 s latency and within 2 m positional error outdoors, 1 m indoors
• Obstacle alerts trigger at least 95 % of the time in my test scenarios
• Audio and haptic feedback are never delivered simultaneously in a way that confuses the user
• Full offline operation for indoor mode, minimal data use outdoors

If this sounds like the kind of challenge you enjoy, outline the toolchain you would use, any similar projects you have tackled, and how quickly you can put together a proof of concept.