Inspiration
Workplace safety inspections are still manual, time-consuming, and prone to human error. Missing a single hazard can lead to accidents, compliance violations, and financial losses. We wanted to build an AI safety auditor that assists inspectors in real time and helps prevent accidents before they happen.
What it does
Sentinel-AI is an AI-powered safety inspection assistant that uses Gemini Live to guide inspectors during inspections. It detects hazards in real time, generates dynamic checklists, collects evidence, asks follow-up questions when needed, verifies compliance, and automatically generates inspection reports.
How we built it
We built Sentinel-AI using Gemini Live, Gemini 3.5 Flash, and Managed Agents. Multiple AI agents collaborate to handle inspection planning, hazard detection, evidence collection, compliance validation, and report generation. The frontend is built with Next.js, while the backend uses FastAPI.
Challenges we ran into
One of the biggest challenges was coordinating multiple AI agents while maintaining inspection context. We also worked on making voice interactions natural and ensuring the AI asks for additional evidence instead of making assumptions when confidence is low.
Accomplishments that we're proud of
- Built a real-time AI safety auditor instead of a static checklist app.
- Created a multi-agent inspection workflow.
- Automated compliance reporting with supporting evidence.
- Enabled natural voice and vision-based inspections using Gemini Live.
What we learned
We learned that AI creates the most value when it augments human expertise rather than replacing it. Combining live vision, voice interaction, and specialized AI agents results in a much more effective inspection experience.
What's next for Sentinel-AI
We plan to add support for smart glasses, IoT sensor integration, predictive risk analysis, offline on-device AI with Gemma, and enterprise EHS platform integrations to make Sentinel-AI a complete workplace safety solution.
Built With
- ai/ml:-google-gemini-api-(gemini-vision)
- automated
- computer-vision-frameworks:-react-(frontend/ui)
- github
- hazard-detection
- image-processing
- node.js-libraries:-react
- npm-other-technologies:-multimodal-ai
- ppe-detection
- prompt-engineering
- react-dom
- real-time-anomaly-detection
- report
- typescript-platforms-&-services:-google-ai-studio-/-gemini-api
Log in or sign up for Devpost to join the conversation.