Problem
Carnaud Metal Boxing Engineering (CMBE) faced the challenge of predicting machine stoppages accurately, which is vital for smooth operations and minimising downtime. In addition, due to IP sensitivities, customers often hesitate to share high-frequency data required for predicting machine stoppages with OEMs. Their concerns about data privacy and misuse of information added complexity to the task. Therefore, they needed effective predictive models that ensure secure data handling.
Solution
CMBE partnered with OctaiPipe to implement a solution centred around generating, deploying, and managing Edge AI capabilities using privacy-preserving Federated Learning Ops. This infrastructure was designed to ensure data privacy and economic profitability for CMBE. It enabled the prediction of machine stoppages without breaching the data privacy of the customers involved.
OctaiPipe’s Impact
OctaiPipe’s implementation yielded substantial economic benefits for CMBE. Creating a digital-as-a-service for CMBE led to reduced spoilage events and increased revenue. In terms of costs, OctaiPipe significantly cut infrastructure and operating costs, while multiplying gross profit percentages and net profit ratios by more than x5.
About OctaiPipe
OctaiPipe is an optimised Federated Edge AI infrastructure that sets a new benchmark for AI deployment to the edge, specifically designed for IoT. OctaiPipe allows data scientists to create Edge AIoT solutions that scale and adapt while reducing the burden of supervision, management, updates, and auditing.
It simplifies the entire ML model lifecycle, from easy setup and training to smooth deployment, fine-tuning, and ongoing learning.
How OctaPipe is Revolutionising Edge AI for IoT:
- Doubles performance in diverse IoT data through continuous learning.
- Provides affordable Edge AI for IoT by lowering training and operational costs.
- Supports continuous learning across 10,000+ devices without data centralisation.
- Ensures resilient AI with less reliance on cloud services and network stability.
- Speeds up AI deployment for up to 500 devices in 3 hours and cuts learning cycles by 90%.
