RepoGuard
MonetizationDependency risk scoring with executive-grade reporting — bridges the gap between raw CVE scanner output and decisions leadership can act on, including SOC 2 evidence.
Security Products, AI Tooling & Infrastructure
Source for these projects is private while they're in development — I'm happy to walk through architecture and code in an interview.
Dependency risk scoring with executive-grade reporting — bridges the gap between raw CVE scanner output and decisions leadership can act on, including SOC 2 evidence.
Compliance automation platform that maps technical controls to NIST CSF, CIS Controls, and SOC 2, continuously collecting evidence instead of relying on point-in-time screenshots.
Runtime drift detection for containers: flags filesystem changes, process spawns, and network connections that deviate from the original image — an early signal of compromise.
Multi-agent orchestration framework: pgvector knowledge store with temporal versioning, typed pub/sub message bus, pre-action conflict resolution between agents, and confidence-decay cascade detection with auto-generated rollback plans.
Self-hosted AI platform for memory management, task orchestration, and homelab monitoring, running local LLMs (Ollama) — designed to extend to family households through lightweight satellite nodes.
Computer-vision game agent research using YOLOv8 detection accelerated with TensorRT, validated through repeatable emulator reliability campaigns. Early alpha — not yet a fully autonomous agent.
A personal learning environment, separate from professional responsibilities — it doesn't claim enterprise SOC ownership or audited compliance.
Baker College case study: designed a small-business network with VLAN segmentation, a DMZ, isolated guest wireless, and passive IDS with centralized logging, plus a maintenance plan covering patch targets, firewall and access reviews, encrypted backups with restore testing, and incident-response roles. A design exercise — targets describe the plan, not measured results.
115+ containers across 7 hosts (Unraid, TrueNAS, Proxmox) managed as code, with VLAN segmentation, a Tailscale mesh, and Cloudflare Tunnels — no inbound ports exposed.
Prometheus, Grafana, Loki, and Alertmanager for metrics and logs; Wazuh SIEM for host detection; CrowdSec for collaborative IP reputation and blocking.
VLAN segmentation, Tailscale, and Cloudflare Zero Trust for network separation and remote access, with Authentik for single sign-on — used in the lab to practice secure administration.
HashiCorp Vault as the single source for infrastructure credentials, with periodic tokens and on-host secret rendering so secrets never land in repositories.
This site's admin dashboard: a read-only FastAPI service queries Prometheus and Loki, published through a Cloudflare Tunnel and verified with Cloudflare Access JWTs.