The 'Running Doom' of AI: Qwen3.5-27B on a 512MB Raspberry Pi Zero 2W

Create an ultra-minimalist execution engine tailored for extremely constrained memory devices (sub-1GB RAM). Focus on optimizing kernel memory footprint and swapping strategies to allow inference on Raspberry Pi Zero class hardware.

Suggested repo: nano-compute

"AI that runs on a potato: Inference for the sub-512MB era."

Estimated effort: 80h