My portfolio isn't a collection of random projects (well.. there is some) — it's a converging architecture.
The real differentiator lives at the intersections, not inside any single cluster.
> The diagram below is a quick visual for technical readers
Security / Detection
YARA rules · PS1 scanner
onion-link detection · OWASP
│
── BRIDGE ──┤── BRIDGE ──
YARA pre-restore │ YARA → prompt injection
malware-aware DR │ pattern defense for LLMs
│
Backup & DR ───────┼─────── Local LLM / RAG
Veeam VBR │ LangGraph · ChromaDB
proxy lifecycle │ BM25+RRF · Ollama
health-check ops │ offline-first
│
── BRIDGE ──┤── BRIDGE ──
VBR telemetry → AI │ topology = policy enforcement
remediation agents │ soul governance · SQLite
│
AI Safety / Agent Governance
PsyClaw · soul drift detection
SHA-256 integrity · MCP server
│
Automation / Sysadmin (substrate)
PowerShell · SCCM · WinRM · VMware
| Repo | What it does | Stack |
|---|---|---|
| PsyClaw | Offline-first RAG agent with LangGraph topology acting as enforceable security policy. Hybrid retrieval (ChromaDB + BM25+RRF), SHA-256 soul drift detection, OWASP-aligned prompt sanitization, and MCP server. Built for memory sovereignty and prompt injection resistance in air-gapped or privacy-sensitive environments. | Python LangGraph ChromaDB FastAPI SQLite |
| AzureAI-CopilotStudio-PersonalAgent-Instructions | Production-tested system instructions for enterprise AI agents. Tiered source hierarchy, anti-hallucination safeguards, and reference implementation patterns (including VeeamGPT). Designed to make agent behavior predictable and auditable at scale. | Azure OpenAI Copilot Studio Prompt Engineering |
| Repo | What it does | Stack |
|---|---|---|
| Veeam-PS1-Scanner-Yara-Rule-Detection-Onion-Links | YARA ruleset + PowerShell scanner that detects Tor .onion C2 infrastructure, BTC/XMR ransomware payment addresses, and common command-and-control patterns inside Veeam restore points. Outputs SIEM-ready JSON and integrates with Secure Restore workflows — turns blind restores into malware-aware, auditable operations. |
PowerShell YARA Veeam Forensics |
| Repo | What it does | Stack |
|---|---|---|
| sccm-veeam-proxy-patching | Graceful Veeam proxy drain during SCCM patch windows. WinRM service management, pre/post task sequencing, and exit code 3010 handling. Prevents backup job failures and data protection gaps during routine maintenance windows. | PowerShell SCCM VMware Veeam |
| Veeam-HealthCheck-Simplifier | CSV/JSON Veeam health check analyzer that produces PowerShell remediation actions plus direct Salesforce/Slack/Teams notifications. Turns hours of manual review into automated, actionable alerts for operations teams. | Python Veeam Analytics |
| Repo | What it does | Stack |
|---|---|---|
| Scrape-n-Email | Lightweight, reliable scraper for news + job digests (RealClearPolitics headlines + Atlanta Craigslist sysadmin roles). MIME-safe email delivery, formula-safe CSV handling, cross-platform logging, and offline testability. Keeps research and opportunity pipelines clean without manual busywork. | Python BeautifulSoup SMTP |
"Backups are operational memory and last-resort truth stores — backup infrastructure is a trustworthy-AI primitive."
This means:
- Veeam is not just a backup tool — it's a ground-truth substrate for AI-assisted recovery decisions.
- YARA is not just malware detection — it's the prompt injection defense layer for AI agents (see
sanitizer.pyin PsyClaw → next evolution:yara-prompt-guard). - PsyClaw is not just orchestration — it's security policy as code (LangGraph Topology = enforcement).
- PsyClaw v1.3 — LangGraph topology rewrite, MCP server, OPA/Rego policy enforcement layer (active development)
Atlanta, GA · Systems Engineer → Infrastructure Engineer → AI Agent Builder · Infrastructure / security engineering roles