Write Once Run Anywhere for GPUs - Portable AI Workloads with Rust and WebAssembly

Explore the potential of Rust and WebAssembly (Wasm) for cross-platform AI workload deployment in this 41-minute conference talk by Michael Yuan from Second State. Learn about LlamaEdge, a lightweight and high-performance LLM inference runtime that leverages WasmEdge to provide a standard WASI-NN API for developers. Discover how this approach enables writing code once and running it on various devices, with WasmEdge handling the translation to native libraries like llama.cpp. Delve into the design and implementation of LlamaEdge, and follow code examples ranging from basic sentence completion to chat bots, RAG agents with vector databases, and Kubernetes-managed applications across heterogeneous clusters. Gain insights into the future of AI application development and deployment in the era of GPUs and cloud computing.

Write Once Run Anywhere, but for GPUs | GPU 时代的“一次编写，到处运行” - Michael Yuan, Second State