LinkEdge: Open-sourced MLOps Integration with IoT Edge
summary
LinkEdge is an open source Edge‑MLOps stack that packages models into Docker containers with an inference toolkit CLI, deploys and manages devices via an Edge manager, and uses a model registry pulled by edge nodes over MQTT. Evaluated on the Scania APS predictive‑maintenance dataset, a deployed model shows inference throughput limits at about 150 requests per second and matches commercial AWS SageMaker/Greengrass performance closely while remaining containerized and vendor‑free.
AI-written summary · about
chapters (9)
- 0:00LinkEdge project intro
- 0:22System architecture
- 2:08Cloud and edge layers
- 4:15CLI implementation
- 6:45Evaluation method
- 9:13Performance results
- 10:58Commercial comparison
- 11:20Conclusion and benefits
- 11:44Q&A
related talks
47mIntroduction to MLOps at the Edge
24mDemo of Kubernetes at the Edge for AI Model Deployments
49mEnable IoT with Edge Computing and Machine Learning
45mAIoT Ops - Using Kubernetes and ML Ops to Build Edge ML Applications
31mBuild Your Intelligent Edge
32mMLOps for edge devices with Amazon SageMaker Edge Manager
29mTeam Teardown: MLOps in the Cloud for IoT
12mUsing AI-enabled Edge Cloud Technologies to Strengthen Europe’s...