← Back to Systems CatalogProprietary Architecture

GozAI

Voice-First Emotional & Accessibility Copilot for Low Vision (Clinically Grounded in Ophthalmic Research)

Visit Live System Commission an Enterprise Build

https://gozai-app.web.app

Production

PLOS ONE & NIH Grounded

Clinical Foundations

Continuous 1 FPS Wayfinding

Multimodal Vision

Sub-500ms Gemini Live

Live Voice Latency

2.2B Low-Vision Individuals

Target Audience

Executive Architectural Overview

Voice-first accessibility companion and spatial navigator bridging the gap between individuals with low vision and inaccessible physical/digital environments. Engineered with clinical foundations as a Doctor of Optometry (O.D.), integrating Gemini Multimodal Live API for continuous 1 FPS spatial hazard detection, prescription medication auditing, audio-haptic wayfinding, and compassionate psychological support.

Core Architectural Accomplishments

Architected voice-first multimodal accessibility copilot for 2.2B low-vision individuals, fusing clinical ocular training (Doctor of Optometry) with Gemini Multimodal Live API over real-time WebSockets.

Engineered continuous 1 FPS camera pipeline detecting physical hazards (stairs, obstacles, spills) and executing live prescription drug verification (Timolol, dosage, expiration).

Built synchronized audio-haptic feedback engine with rising-tone offline light-metering to assist spatial orientation in unfamiliar environments.

Deployed on Google Cloud Run and Firebase Firestore with brutalist accessibility standards exceeding WCAG AAA compliance.

Deterministic Guarantees

Concurrency Model

Sub-500ms bidirectional PCM audio streaming, ephemeral token isolation, zero data retention for HIPAA/EHR privacy.

AI & Protocol Layer

Gemini Multimodal Live API (2.0 Flash), 1 FPS camera vision pipeline, clinical empathy persona.

Database & Infrastructure

Flutter 3.47 (Web & Mobile), Google Cloud Run microservices, Firebase Firestore realtime store.

SUBSYSTEMS • ARCHITECTURAL MODULES

Integrated System Components

MODULE 01

Clinical Foundation & Medication Safety Engine

Real-time multimodal auditing of prescription pill bottles (identifies active drug name, dosage instructions, and expiration dates) to prevent life-threatening medication errors.

MODULE 02

Continuous 1 FPS Spatial Wayfinding & Hazard Detection

Hands-free camera stream analyzing walking paths in real time to alert users of low-contrast stairs, wet floors, and overhanging obstacles using synchronized vibro-acoustic feedback.

MODULE 03

Voice-First Clinical Empathy Persona

Low-latency conversational audio companion calibrated to counteract the cognitive fatigue, isolation, and anxiety associated with progressive vision loss.

MODULE 04

Brutalist Accessibility UI & Offline Light-Meter

Ultra-high-contrast interface with screen-reader-first architecture, accompanied by an offline ambient light sensor that emits rising melodic audio tones toward natural light sources.

MODULE 05

Visual-Only Digital UI Navigation

Interprets inaccessible visual elements, untagged buttons, and complex image-based digital workflows without manual screenshot taking.

PRODUCTION STACK • VERIFIED

Technology Architecture Matrix

Frontend & Runtime

Flutter 3.47
High-Contrast Brutalist Accessibility UI
Vibro-Acoustic Haptic Feedback
WebAudio Tone Synthesizer
Continuous 1 FPS Camera Pipeline

DB & Concurrency

Google Cloud Run
Firebase Firestore
Sub-500ms Audio Streaming Channels
Ephemeral Session Token Vault
Offline Light-Meter Calibration

AI & Realtime

Gemini Multimodal Live API (2.0 Flash)
Clinical Empathy Audio Persona
Medication Safety OCR (Pill/Expiration Parsing)
Real-Time Hazard Detection
1 FPS Spatial Wayfinding

Payments & Protocols

Open Clinical Protocol
FHIR / Medical Record Compatible Data Flow
Zero-Tracking Privacy Guard
WebSockets Audio Streaming

Previous Case Study

Huiyi

Commission an Enterprise Platform

Next Case Study

CaelumOS