We'll define a system architecture that brings modern AI to out-of-band defensive cybersecurity monitoring on edge deployments in a size, weight, and power (SWaP)-efficient form factor. Many edge networks are in regions with denied, degraded, intermittent, limited internet access and can't rely on scale-out compute in cloud or remote data center. These networks are further constrained by SWaP available in edge computing locations. This architecture will describe a network test access point system feeding a defensive security enclave that deploys a suite of monitoring tools. We demonstrate the convergence of the Ethernet fabric and Packet Broker on a single Spectrum switch that load balances network traffic to an array of converged servers powered by BlueField DPUs and NVIDIA Morpheus for GPU-accelerated, AI-enabled network detection and response.