How Tata Elxsi detects industrial safety risks in seconds on AWS
aws ·
Tata Elxsi has developed IRIS, a real-time industrial safety platform built on AWS that detects safety risks in seconds rather than minutes. The system utilizes edge computing with AWS IoT Greengrass to filter video frames locally, sending only metadata and relevant images to the cloud to reduce bandwidth usage by up to 80 percent. Amazon Kinesis Data Streams serves as the event backbone, routing structured data to Amazon SageMaker AI for computer vision inference. Custom models analyze footage for PPE compliance, restricted zone intrusions, and worker proximity risks using GPU-accelerated instances. A correlation layer implemented via AWS Lambda and DynamoDB reduces false positives by evaluating temporal patterns before generating alerts. Automated escalation workflows ensure critical incidents are notified within seconds through dashboards, SMS, and email. Production results indicate a 15–20% reduction in recordable safety incidents and a 30% decrease in surveillance operating costs. The architecture supports continuous learning through active loops, ensuring model accuracy improves over time based on real-world production data.
aws