AI infrastructure startup focused on reducing enterprise AI inference costs through KV cache optimization.