Research Computing Infrastructure

Dr. Martin designs, builds, and operates the self-hosted computing infrastructure behind his security research, AI inference, and expert witness work. Its purpose is twofold: to provide isolated, reproducible environments in which confidential code and case data can be analyzed entirely under his own control, and to supply the capacity to run compute-intensive analysis at scale. In practice that means rapid provisioning of segmented, multi-OS test networks; controlled handling of sensitive material; and experiments that can be torn down and repeated exactly.

Server rack with virtualization host, hot-swap drive bays, and CyberPower UPS
The mobile rack: virtualization host with hot-swap storage bays, KVM, and battery-backed power.

Consolidated virtualization platform

The environment runs on Proxmox, using PCIe passthrough to give dedicated virtual machines direct control of physical hardware — a firewall VM (OPNsense) owns its network interfaces, and a storage VM (TrueNAS) owns its disk controllers. This consolidates what was previously a rack of separate physical appliances into a single chassis without giving up the isolation or performance of dedicated hardware. The same consolidation is what made room for new capability: the platform shares a chassis with the LLM inference testbed used for local analysis of frontier-class language models.

Storage and data integrity

Primary storage is high-capacity ZFS with end-to-end data-integrity verification, snapshots, and replication. Checksumming at every layer means silent corruption is detected rather than propagated — a property that matters when the data at rest includes forensic images and case material that must remain demonstrably intact. Hot-swap drive cages keep pool maintenance and capacity changes routine.

Server chassis interior showing Noctua CPU cooler, Corsair RM850x power supply, Intel Arc Pro GPU, and dual hot-swap drive cages with routed SATA cabling
Inside a storage and virtualization host: dual hot-swap drive cages with individually routed and labeled SATA runs, Noctua cooling, and an Intel Arc Pro GPU.

Network segmentation and test networks

Custom-built OpenBSD and OPNsense network appliances provide deep packet inspection, traffic shaping, and segmentation. Segmentation is what turns the infrastructure into a research instrument: suspect firmware, malware samples, and multi-OS test environments run on isolated network segments that can be created for an investigation and destroyed when it ends, with sensitive material never crossing onto general-purpose networks.

Availability

The rack runs on conditioned, battery-backed power, and storage replication provides recovery points for the systems and data behind active casework — capability stays available, under Dr. Martin’s control, for the duration of a confidential investigation.