Machine Learning in Fastly's Compute@Edge

Explore the integration of machine learning models in Fastly's Compute@Edge environment through WebAssembly modules and wasi-nn in this conference talk. Discover the advancements made to enable efficient execution in a stateless FaaS environment, including extensions to the wasi-nn spec, revisions to host APIs, security-related tradeoff considerations, and the introduction of a new proxy backend based on the KServe protocol. Witness a demonstration of these functionalities through a Compute@Edge service utilizing OpenVINO, ONNX, and PyTorch for classification and generative AI applications.

Machine Learning in Fastly's Compute@Edge - Andrew Brown, Intel & Matthew Tamayo-Rios, Fastly