Tensormesh raises $4.5M to squeeze more inference out of AI server loads October 23, 2025 by kamal Tensormesh uses an expanded form of KV Caching to make inference loads as much as ten times more efficient.