Orchestrating Machine Learning on Edge Devices Using Akri and WebAssembly
summary
Combines Project Akri and WebAssembly to discover cameras/phones as Kubernetes resources on a k3s cluster and run ML inference via a wasm shim and Akri agent/controller. Demonstrates MobileNeRF rendered 3D output on an Android phone discovered by Akri, using AOT-compiled WASM and warm‑up to cut binary size to ~3.3 MB and speed up MobileNet-style workloads by roughly 40% (and large inference paths by over 100% in some cases). The pipeline is implemented and the demo works.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:04What is WebAssembly
- 5:14Why WebAssembly excels
- 6:59Introducing Project Akri
- 9:48Akri with WebAssembly
- 13:32Demo setup for ML inference
- 14:37Optimizing WebAssembly binaries
- 19:37Live NeRF demo
- 23:58Best practices and conclusion
related talks
12mLightning Talk: Orchestrating Machine Learning on Edge Devices...
21mConnecting to devices at the edge with minimal footprint using AKS Edge Essentials and Akri on WASMs
29mLiving on the Edge: Using IoT Devices on Kubernetes WebAssembly Applications
24mFrom Sensors to Servers: Efficient Edge Computing with Akri and WebAssembly
33mWebAssembly based AI as a Service with Kubernetes
32mWebAssembly Based AI as a Service on the Edge with Kubernetes
36mApplied AI on the edge
36mWebAssembly-based Edge AI Sensing Pipeline Orches...