Skip to content


EdgeSage is our Edge AI software stack that is built on open-source components like Kubernetes, Kubeflow

Building an Edge AI software stack from open-source components like Kubernetes and Kubeflow is a great way to harness the power of container orchestration and machine learning on edge devices. Here’s a high-level overview of the steps used to create such a stack:

  1. Choose Hardware to run EdgeSage:
    • First, select the edge devices or hardware platforms that you plan to deploy your Edge AI software on. Ensure that they have the necessary processing power and compatibility with Kubernetes. Edge servers are specialized computing nodes that are strategically placed at the edge of a network, closer to the data sources and end-users. These servers are designed to process and manage data in real-time, offering a more responsive and efficient approach compared to traditional cloud computing, which involves data centers located at a considerable distance from the end-users. By processing data closer to where it’s generated, edge servers reduce latency, enhance security, and ensure faster response times.
  2. Kubernetes:
    • EdgeSage installs Kubernetes on your edge devices. You can use lightweight Kubernetes distributions like K3s or MicroK8s, which are designed for resource-constrained environments.
  3. Containerization:
    • Containerizes your AI models and applications using Docker. This ensures that your workloads are portable and can run on any Kubernetes cluster.
  4. KubeFlow:
    • Installs KubeFlow on your Kubernetes cluster. KubeFlow provides tools and libraries for deploying, monitoring, and managing ML workloads on Kubernetes.
  5. Model Training and Deployment:
    • Uses KubeFlow Pipelines or other tools to create, train, and package your machine learning models.
  6. Model Serving:
    • Deploys your trained models as Kubernetes Deployments or StatefulSets using KubeFlow Serving or other model serving tools.
  7. EdgeSage Device Integration:
    • Ensures that your exiting edge devices can connect to the EdgeSage cluster. This involves setting up network configurations and ensuring connectivity.
  8. EdgeSage Inference Optimization:
    • Optimizes your AI models for edge inference. Edge devices typically have resource constraints, so you might need to use model quantization and other techniques to reduce model size and inference latency.
  9. EdgeSage Device Management:
    • Implements device management and monitoring for your edge devices. This can be done using Kubernetes operators, custom scripts, or other open-source tools.
  10. Security:
    • Pay special attention to security, as edge devices can be vulnerable. Uses Kubernetes security best practices, implement network security measures, and ensure that your models and data are protected.
  11. Scaling and Load Balancing:
    • Implements scaling and load balancing as needed. Kubernetes offers these features out of the box.
  12. Monitoring and Logging:
    • Uses Kubernetes-native monitoring and logging solutions like Prometheus and Grafana to keep an eye on the performance of your Edge AI stack.
  13. Updates and Maintenance:
    • Regularly updates and maintains your Edge AI software stack to ensure that it’s running smoothly and securely.

Let’s talk about your next AI project

Ready to talk