>rig for local inference for a home agent (powered by four RTX PRO 6000 Blackwell Max-Qs)

can you elaborate at all on what sort of rig you went with, beyond the big $$ GPUs?