Lightning Talk: Scaling AI With Wasm and Edge Computing
Lightningat KubeCon + CloudNativeCon China 2025 · on Tuesday, 2025-06-10· at 17:08· Level 16 | Grand Ballroom I·
summary
WasmEdge presents a Wasm-based, OpenAI-compatible API server that runs LLMs and vision/voice models in a sandbox across devices including NVIDIA GPUs and Macs, using runtimes like llama.cpp and PyTorch for portability. Demonstrated apps include AI streamers, multilingual video dubbing, and a real-time conversational agent running on a ~$15 edge device. Claims operational scale: the Gaia network runs on 600k machines and an e‑commerce deployment processes 500k videos/day, showing low-latency, high-throughput edge inference.
AI-written summary · about
chapters (9)
- 0:00WasmMatch and AI scaling
- 1:03Current AI architecture limits
- 2:33Why WebAssembly for AI
- 3:19WasmMatch API server features
- 4:24AI streamer demo
- 6:15Video translation tool
- 7:20Conversational agent demo
- 8:44Scalability proof and contributors
- 9:27Closing
related talks
48mWorkshop: Efficient and Portable AI / LLM Inference on the Edge Cloud
07mLightning Talk: AI at the Edge: ONNX Inference in WASM on Featherweight K0s
25mLeveraging Wasm for Portable AI Inference Across GPUs, CPUs, OS & Cloud-Nativ...
37mGetting Started with AI and WebAssembly
41mCreating Versatile AI Agents Through Wasm+Rust
43mBuilding Serverless AI Workflows with Wasm+Rust
31mRunning AI workloads with WASM - is it production ready? [Panel]
12mLightning Talk: Orchestrating Machine Learning on Edge Devices...