1 points | by openfluke 8 hours ago ago
1 comments
I’ve been building Paragon, a GPU-agnostic AI framework in Go with WebGPU (browser) and Vulkan (native) acceleration. It runs without CUDA lock-in and shares the same neural network code across frontend and backend.
Showcase (MNIST cross-bench): https://github.com/openfluke/Paragon-MNIST-CrossBench Framework repo: https://github.com/openfluke/paragon POC demo video (physics sandbox): https://www.youtube.com/watch?v=1pF2x3rs2gg
More details + docs: https://openfluke.com
Would love feedback from the Go, ML, and systems communities.
I’ve been building Paragon, a GPU-agnostic AI framework in Go with WebGPU (browser) and Vulkan (native) acceleration. It runs without CUDA lock-in and shares the same neural network code across frontend and backend.
Showcase (MNIST cross-bench): https://github.com/openfluke/Paragon-MNIST-CrossBench Framework repo: https://github.com/openfluke/paragon POC demo video (physics sandbox): https://www.youtube.com/watch?v=1pF2x3rs2gg
More details + docs: https://openfluke.com
Would love feedback from the Go, ML, and systems communities.