A major construction company operating one of the largest regional sites approached us to help reduce rising incident rates tied to poor communication. Key issues included inconsistent safety messaging, reliance on hand signals, verbal instructions, and frequent misunderstandings due to language barriers on multilingual crews.
The client sought a practical, field-tested solution that could centralize communication, deliver safety information in multiple languages, and operate reliably in low-connectivity environments with limited hardware.

We conducted field interviews and site visits to understand real worker behavior and communication habits. Based on this, we designed and delivered a mobile-first solution focused on simplicity, accessibility, and multilingual support.

We developed a task-specific AI assistant trained on real-world construction safety documentation using open-source NLP frameworks (spaCy, Rasa NLU). The bot answered worker queries, delivered illustrated safety PDFs, and shared short video explainers. A feedback loop allowed ongoing content improvement based on user ratings and comments.

A built-in communication layer allowed users to send alerts, ask questions, and notify supervisors about safety risks. Features included:
- Group and individual messaging
- Offline message queuing
- Compatibility with low-spec Android devices
- Abuse reporting and user blocking for moderation
- Learning & Engagement Features
The app included lightweight tracking tools to help workers follow key safety practices and complete short learning paths. These were optional and intended to support compliance or internal recognition goals — not formal certification.

To address communication gaps, the app supported seven languages — English, Hindi, Urdu, Arabic, Bengali, Malayalam, and Mandarin — across both chatbot responses and UI, including support for RTL scripts.

The platform was integrated with Rocket. Chat for messaging and Keycloak for secure user authentication and role management, ensuring enterprise-level scalability and modularity.