Inspiration

Workplace ergonomics injuries and driver fatigue cost industries billions annually, while traditional OHS compliance remains reactive, paper-heavy, and manual. ErgoSafe Reborn V3 was inspired by the need for an active, autonomous safety wingman that prevents injuries before they occur. By combining real-time biomechanical strain detection with zero-knowledge privacy standards, we set out to build a platform that protects workers' health without compromising their personal data rights.

What it does

ErgoSafe Reborn V3 acts as an integrated OHS Command Centre powered by specialized AI agents:

Ergonomics Engine: Monitors real-time C1–C7 cervical spine tilt and trapezius strain to trigger live micro-stretch coaching alerts. Driver Safety Telemetry: Tracks shift hours, reaction time drops, and fatigue scoring via /api/v1/fatigue-score to enforce statutory rest lockouts. Governance & POPI Handshake: Uses a zero-knowledge encryption layer to verify employee digital identity and pre-qualifications without storing invasive surveillance data. OHS Legal Shield: Generates statutory Section 8(1) and Section 37/38 compliance proof and audit-ready legal dossiers.

How we built it

Frontend: React, TypeScript, and Tailwind CSS optimized for zero layout shift and executive command metrics. AI Architecture: Multi-agent framework built with Google Cloud Vertex AI / Gemini tools and Antigravity IDE for autonomous decision orchestration. Audio & Media: Synchronized interactive Web Audio engine featuring dynamic audio briefs alongside 1080p narrated walkthroughs. Deployment & CI/CD: Edge-deployed via Vercel with automated build validation pipelines.

Challenges we ran into

Balancing continuous telemetry with worker privacy required building a strictly client-side zero-knowledge consent architecture. Additionally, fine-tuning the dynamic Web Audio synchronization with real-time UI telemetry states while preserving 60 FPS frontend performance called for precise state management and CSS layout fixes.

Accomplishments that we're proud of

Built an end-to-end multi-agent OHS platform that bridges statutory South African labor law with modern AI ergonomics. Designed a zero-knowledge privacy handshake ensuring compliance data stays secure. Achieved zero layout overflow across complex telemetry screens.

What we learned

Best practices for streaming micro-telemetry payloads to AI agents without incurring frontend latency. Effective strategies for structuring multi-agent workflows to handle concurrent safety alerts, fatigue scoring, and legal audit generation.

What's next for ErgoSafe Reborn V3

On-device edge processing for offline heavy machinery operator monitoring. Expansion into real-time hazard detection using local camera feeds at service station and warehouse facilities.

Pre-Existing Code & Framework Disclosure

Per hackathon rules, the following pre-existing frameworks and baseline modules were incorporated:

  1. Core Architectural Framework: Baseline UI layout, statutory regulatory schemas (SA OHS Act 85, R638, SANS standards), and standard React/Vite boilerplate were leveraged from existing OHS foundation templates.
  2. Hackathon Incremental Build: All agentic workflows, multi-agent CrewAI orchestration, GEAR alignment, T=0.0 deterministic guardrails, Google Cloud Vertex AI integrations, and mobile responsiveness refactoring were newly authored, built, and integrated during the hackathon submission window.
  3. Standard Tooling & AI Assistance: Standard development utilities, open-source npm libraries, and AI coding assistants were utilized during development as permitted by official rules.

Built With

  • anti-gravity-ide
  • gemini-3.5-flash
  • gemma-4
  • github
  • google-cloud-firestore
  • google-cloud-pubsub
  • google-cloud-run
  • google-genai-sdk
  • react
  • replit
  • tailwind-css
  • typescript
  • vercel
  • vs-code
Share this project:

Updates

Submission history