Running AI workloads with WASM - is it production ready? [Panel]
summary
Wasm is framed as a practical inference layer: a write-once run-anywhere runtime that abstracts GPU drivers and model formats, delivers compact footprint, near-native performance, and cross-platform compatibility, and enforces sandboxed isolation. Concrete demos include a Rust-based pipeline handling 700,000 videos per day for transcription, deployments using SpinCube on Kubernetes to attach GPUs without changing infra, and edge/browser runs like LlamaEdge on Raspberry Pi and WebLLM+WebGPU running 1–2B models locally with low latency.
AI-written summary · about
chapters (10)
- 0:00Panel intro on WASM and AI
- 1:01Generative AI landscape trends
- 2:26AI ecosystem evolution
- 7:48AI programming languages
- 12:19WebAssembly in generative AI
- 15:52Why WASM for AI inference
- 18:22WASM in production AI
- 22:38Kubernetes with WASM for AI
- 25:49Browser AI with WASM
- 29:59Closing & Q&A
related talks
25mLeveraging Wasm for Portable AI Inference Across GPUs, CPUs, OS & Cloud-Nativ...
01h 04mDevelop, Build & Deploy Lightweight AI Services with WebAssembly
40mWebAssembly for AI Infra : A Lightweight, Fast, and Secure Alternative Approach
37mGetting Started with AI and WebAssembly
32mWebAssembly Based AI as a Service on the Edge with Kubernetes
27mWasm Is Becoming the Runtime for LLMs
33mWebAssembly based AI as a Service with Kubernetes
41mWebAssembly Based AI as a Service