
Sign up to save your podcasts
Or


In our first episode, AMD’s Usman Pirzada turns the microphone around and interviews Strix Halo Stars host Donato Capitella about his journey into local AI.
Donato shares how experimenting with Strix Halo led to the Toolboxes project, repeatable benchmarks and a community of developers building specialised inference engines. They discuss Vulkan, ROCm, llama.cpp, Helogen, Gufo and ROCmFPX; how models from Qwen, DeepSeek and GLM have changed what runs locally; and what comes next, from Gorgon Halo to combining local models with cloud advisors.
This episode is sponsored by AMD.
Chapters
00:00 Welcome and Donato’s Strix Halo journey
09:50 Building and sharing Toolboxes
17:38 Community engines and optimisations
26:51 Local models: Qwen, DeepSeek and GLM
37:20 What’s next for local AI?
Strix Halo Toolboxes: https://strix-halo-toolboxes.com/
Show website: https://strix-halo-stars.com/