Ergo-Safe Reborn (OHS Haven)

Multi-Agent Autonomy for Global Workplace Safety & Compliance.

Comment

Inspiration

The inspiration for Ergo-Safe Reborn evolved from a critical observation during my time in high-stakes environments like Chevron: OHS fails not because of a lack of rules, but because of a lack of stewardship. Most safety hazards become "white elephants"—reported but ignored—due to the massive administrative burden required to track them. We realized that for a workplace to be truly safe, accountability must be autonomous. We didn't just want to build a better manual; we wanted to build a Stewardship & Escalation Engine. By combining Admin-Zero Automation with Gemini 3, we’ve created a system that removes the paperwork and forces awareness. If a hazard isn't addressed, the system autonomously escalates it up the chain—from supervisor to CEO—ensuring that safety is a living part of the organizational culture, not just a line in a file.

What it does

Ergo-Safe Reborn is an autonomous OHS Stewardship platform that removes the administrative burden of safety compliance while enforcing organizational accountability. OHS Stewardship & Escalation Engine: A proactive accountability matrix that ensures no hazard is ignored. Using a "Steel Thread" logic, the system autonomously escalates unaddressed risks: Day 5 (Friday): Level 1 alert to the Direct Supervisor. Day 8 (Monday): Level 2 escalation to the Department Head/Leader. Day 14: Executive Alert to the CEO Dashboard, flagging the department for "High Liability Risk." Admin-Zero Automation: We eliminated the OHS "paperwork mountain." The system automatically files signed-off assessments into digital employee files and generates the monthly Compliance & Risk Exposure Report for Board of Director review without human intervention. CEO Safety Pulse Dashboard: A high-level leadership view featuring a single "Company Safety Pulse" dial. This reflects real-time OHS Act adherence across the entire organization and includes a "Stop Work" indicator for departments falling below critical compliance thresholds (60%). Multilingual Melly AI: A world-first safety copilot supporting English, Zulu, Xhosa, Spanish, and Mandarin. Melly provides real-time voice guidance and localized "Legal Shock" warnings to ensure every worker and manager understands their personal liability under the Law. 3D Physics Training: 20+ premium interactive cards powered by the Anti-Gravity engine, providing scientifically accurate ergonomic instruction to prevent MSDs (Musculoskeletal Disorders).

How we built it

We built a "Headless" production pipeline to ensure we could scale content without manual bottlenecks: LLM Reasoning: We utilized Gemini 3 to power the "Agent Reasoning Log," allowing our AI agents to process complex safety acts and regulatory data. Backend: Hosted on Google Cloud Console, utilizing Codex to translate natural language prompts into animation logic. Frontend: A high-performance dashboard deployed via Vercel for seamless user interaction. Visuals: High-fidelity 3D assets rendered through the Anti-Gravity physics engine to ensure ergonomic demonstrations are scientifically accurate.

Challenges we ran into

The biggest challenge was "Copyright Kak" and logo blurring in our automated video feeds, particularly for our sports and high-volume segments. We solved this by building custom computer-vision filters to automatically protect intellectual property while maintaining the speed of the AI pipeline. Additionally, syncing multilingual audio (Zulu, Xhosa, Sotho) with real-time agent lip-sync required deep fine-tuning of our NLP models.

Accomplishments that we're proud of

Developed and Upscaled ErgoSafe for OHS Haven: Successfully built a market-ready platform focused on office ergonomics compliance, self-assessments, and high-level HR reporting. Engineered a Multi-Agent Ecosystem: Deployed a functional backend where specialized agents like Sarah, Marcus, and Melly collaborate on real-time safety indexing and risk calibration. Created High-Fidelity Multilingual Assets: Directed the production of high-quality animated videos, including a cheerful children's educational series ("Iris & Friends") and professional enterprise demos featuring Melly, the OHS Haven AI assistant.hysics-Engine Integration: Successfully integrated the Anti-Gravity physics engine to render lifelike movements and ergonomic demonstrations for our agents. Automated Pipeline: Built a "zero-touch" content pipeline where Anti-Gravity renders are automatically processed and branded for the OHS Haven dashboard. Localized Global Content: Integrated specific South African English voiceovers and localized branding to ensure the platform resonates with both corporate and community-level users.

What we learned

nfrastructure First ("Steel Thread"): We learned the critical importance of stabilizing core infrastructure and backend logic before layering on complex features like lip-synced avatars for deployment. Enterprise-Grade Analytics: Realized that to lead the market, a platform must prioritize advanced features like Predictive Analytics and Gamification to keep users engaged with safety protocols. Niche-Specific Branding: Learned how to effectively pivot visual styles and tones between different target audiences—maintaining a "Cocomelon-style" for kids while delivering a sleek, authoritative "Safe-Tech" vibe for corporate clients. Scalability Through Automation: Discovered that using a headless production pipeline (Google Cloud + Codex) is the only way to meet the demand for high-volume, localized educational content without losing quality.The Power of Physics in AI: We learned that using a dedicated physics engine like Anti-Gravity provides a level of "visual truth" that standard generative AI still struggles with, especially for safety training where physical posture is everything. Cloud Orchestration: We learned how to balance the heavy GPU requirements of Anti-Gravity within our Google Cloud environment to ensure fast, scalable video delivery.

What's next for Ergo-Safe Reborn (OHS Haven)

We are expanding into a full Mini-LMS (Learning Management System) that is globally compliant and fully automated. Our next milestone is integrating real-time wearable data so our agents can provide "on-the-spot" ergonomic corrections to workers in the field.

Built With

Share this project:

Updates

posted an update —

Milestone Reached: Production Deployment & GEAR-Verified Architecture Sync. Since the initial submission, Ergo-Safe Reborn has evolved from a hackathon concept into a production-ready compliance engine. I’ve focused this iteration on strengthening the "Pillars of Integrity" that define industrial safety.

Key Updates & New Features:

Production Launch: The app is now live at ergo-safe-reborn.vercel.app. This move to Vercel ensures global low-latency access for safety officers in the field.

GEAR Framework Alignment: I’ve fully synchronized the codebase with the GEAR Framework (Reasoning, Grounding, Tool Use, Security).

Melly AI Enhanced: Our AI assistant, Melly, now features deeper Reasoning loops to explain complex OHS regulatory changes in real-time, moving beyond simple Q&A.

Regulatory Grounding: Strengthened the grounding rules for the South African OHS Act 85 of 1993, ensuring that every safety recommendation is legally substantiated.

Security First: Integrated strict PII masking and data tainting to ensure corporate safety records remain confidential and secure.

Behind the Code: The project now runs on a modernized Next.js and Tailwind CSS stack, orchestrated via Google Antigravity. This transition allows for faster iteration and seamless updates as safety regulations evolve.

What’s Next? I am currently working on a "Safety Literacy" bridge between Ergo-Safe and our sister project, Play 2 Learn English, to bring safety awareness to the next generation of workers.

Log in or sign up for Devpost to join the conversation.