International Journal of Scientific Research and Engineering Development

Artificial intelligence (AI) is increasingly used in occupational safety through computer vision, wearable sensing, Internet-of-Things (IoT) networks, equipment telemetry, and predictive analytics. Yet the literature remains fragmented across sensing modalities and frequently evaluates algorithmic accuracy without establishing whether AI deployment improves worker safety outcomes. This structured technical review synthesizes recent evidence from 2018–2026 and proposes a quality- and uncertainty-aware multimodal framework that connects sensing, edge processing, feature fusion, risk inference, selective decision-making, intervention, and governance. Recent reviews show rapid growth in PPE detection, worker-state monitoring, construction IoT, fatigue assessment, and accident prediction, but external validation, calibration, real-world robustness, and prospective outcome evaluation remain limited. To distinguish technological performance from safety effectiveness, a five-level evidence-maturity model is introduced: measurement validity (E1), algorithmic validity (E2), operational validity (E3), intervention effectiveness (E4), and worker health/safety outcomes (E5). A selective decision rule further distinguishes alert, no-alert, and indeterminate states when sensing quality or predictive confidence is insufficient. The framework links AI outputs to the hierarchy of controls and treats privacy, cybersecurity, worker autonomy, and human oversight as system-level requirements. The review concludes that AI can strengthen occupational risk detection and decision support, but high model performance should not be interpreted as evidence of reduced injury or illness without prospective workplace validation.

Shifa Bilal Tamboli, Simeen Phiroj Mulani, Arman Tajuddin Shiakh,"Multimodal AI for Occupational Safety: A Quality-Aware Framework for Sensing, Risk Inference, Intervention, and Governance" International Journal of Scientific Research and Engineering Development, V9(2): Page(3849-3859) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.