Sitegaurd AI

Inspiration

Construction continues to be one of the most dangerous industries in the world with thousands of preventable injuries and deaths each year. We found that existing safety systems are predominantly reactive, detecting incidents after they happen. We’re trying to figure out how AI & wearable tech can make workplace safety a preventative measure.

What it does

SiteGuard AI is an AI-powered safety assistant for Meta Smart Glasses. It continuously scans the worker’s environment, reads the context of the environment and alerts the worker in real-time through voice for possible risks such as missing PPE, fall hazards, electrical hazards, overhead loads, machine hazards, restricted areas etc. It also can notify supervisors with contextual information for critical situations.

How we built it

We built the system on a multimodal AI pipeline comprising computer vision, speech recognition, retrieval-augmented generation (RAG) and voice feedback. The aim is to combine knowledge of equipment, knowledge of site-specific hazards and safety regulations to make context-aware decisions rather than relying purely on object detection. Also, we created a supervisor workflow to configure site-specific hazards before each shift.

Challenges we ran into

One of the biggest challanges was to design an AI system that understands context, not only objects. The construction site is a dynamic environment where hazards are dependent on several factors including worker position, personal protective equipment, machinery and site-specific conditions. We also had to weigh technical feasibility versus a hands-free user experience for real construction workers.

Accomplishments that we're proud of

We developed a holistic AI safety concept that integrates wearable computing, real-time hazard detection, voice interaction and supervisor intelligence in a single system. We think our layered approach to contextual safety makes the solution more practical than traditional monitoring systems.

What we learned

This project helped us better understand multimodal AI, computer vision, retrieval-augmented generation, human-centered design, and the practical challenges of deploying AI in safety-critical environments. We also learned the importance of designing technology around real user workflows instead of simply adding AI features.

What's next for Sitegaurd AI

Our next steps include developing a working prototype on wearable hardware, integrating real-time computer vision models, evaluating the system in realistic construction scenarios, and improving detection accuracy while maintaining low latency for on-site deployment.