Back to Projects
Safesight AI
A cloud + edge system that detects hazards (slips, spills, equipment faults) from live camera feeds, notifies stakeholders automatically, and provides an incident dashboard for analytics and compliance.
Overview
Built SafeSight AI, a cloud+edge incident detection MVP (React/TypeScript, Node.js, Kubernetes, PyTorch + Triton) that reduced manual incident reporting by automating real-time detection, clip extraction, and multi-step remediation workflows implemented edge inference, event-sourcing workflows (Temporal), and an analytics dashboard for operational insights.
Key Achievements
Developed an AI-powered workplace safety platform to automate hazard detection and incident reporting workflows
Implemented a containerized microservices architecture using React, Node.js, Python, and Docker for scalable deployment
Automated build, linting, testing, and deployment validation through GitHub Actions CI/CD pipelines
Designed a real-time hazard monitoring framework capable of detecting slips, spills, and falls from video streams
Built RESTful APIs and database schemas to manage incident reports, hazard history, and safety analytics
Reduced deployment inconsistencies by containerizing frontend, backend, and ML services with Docker Compose
Established a production-ready development workflow with automated code quality checks and testing
Laid the groundwork for integrating YOLOv8/Hugging Face computer vision models into a workplace safety monitoring system
Tech Stack
React.jsViteNode.jsExpress.jsPythonPostgreSQLDockerGitHub ActionsYOLOv8JestESLintDocker Compose