DOUG CASHIO — ENTERPRISE AI & CYBERSECURITY SOLUTIONS ARCHITECT

Starship Construction & Modification Timeline

The following chronicle documents the ZeusApollo fleet's evolution — from first warp core ignition to current multi-vector operations. Each milestone is a new system coming online aboard the Enterprise, dramatically expanding our operational ceiling. This is not a hobby project. This is a living proof-of-concept for enterprise-grade sovereign AI.

Bridge Crew Configuration — Multi-Agent Orchestration

On the Enterprise, Captain Picard doesn't personally fire torpedoes, scan anomalies, or reroute plasma conduits. He delegates to specialized officers who each excel at their station. The ZeusApollo "Picard Routine" follows the same doctrine: each AI agent is assigned a bridge station based on its unique strengths, and the Admiral orchestrates them as a unified, battle-ready crew.

"Shields up, Red Alert" = Wazuh SIEM ingesting and correlating threat telemetry across all nodes. "Sensors scanning" = Kali Linux running AI-assisted Nmap reconnaissance. "Engineering, I need more power" = Ollama inference shifted to the RTX 3080 for 16× throughput gains.

This is not role-play. This is a production-grade, cost-optimized, multi-model AI orchestration architecture operated by a single human commanding officer.

Operational Capabilities & Efficiency Gains

GPU Migration (The Warp Core Upgrade): Phase 2 transitions AI inference from Zeus's CPU (~5 tok/s — impulse speed) to the GPU workstation's RTX 3080 (~80 tok/s — warp factor 8). This is the single largest performance gain in the fleet's history: a 16× throughput improvement that transforms local AI from a novelty into a production-grade enterprise capability.

Automated Deployment (Replicator Technology): Manual VM/container creation replaced by scripted deployments. The Kali Linux AI script automates the entire process — download, create, configure, and boot — bypassing 30–60 minutes of manual Proxmox clicking. Cloud-Init and Preseed templates make new deployments as frictionless as replicating Earl Grey.

Proxmox Backup Automation: All nodes (VM 100, CT 101–103) back up automatically at 21:00 nightly to genesis-nfs. No manual intervention. The fleet protects itself.

MCP Integration (Universal Translator): Custom MCP servers bridge the gap between AI agents and infrastructure. The AdGuard MCP lets Claude query DNS statistics in natural language. The Apollo health monitor MCP provides live CPU/RAM/temperature telemetry. Future MCP servers for Wazuh and Home Assistant will complete the automation loop.

Estimated OpEx — Running the Fleet

Network Topology — Homelab Architecture

The fleet's command structure mapped across the homelab network. Every node is a starship in the task force, interconnected via the main deflector (router/switch) with AdGuard Home filtering all subspace communications. Drag nodes to explore. Hover for telemetry.

Capability Score by Phase — Fleet Power Progression

Each phase represents a quantum leap in operational capability. Scores are composite ratings across AI inference speed, automation coverage, security posture, and operational resilience. Phase 0 is baseline manual operation; Phase 4 is the projected end-state with full HA, Wazuh SIEM, and GPU-accelerated inference at scale.

Local vs. Cloud — Cost Trajectory Over 36 Months

The crossover point where on-premises investment pays for itself against equivalent cloud GPU rental. Green line: cumulative local cost (hardware CapEx + monthly OpEx). Red lines: equivalent AWS/Azure instance hours for comparable AI workloads at 8 hours/day. The math is decisive.

Status Board — Active & Deferred Items

Strategic Vision — Summer 2026 End-State

By the end of summer 2026, the ZeusApollo fleet achieves full operational autonomy across five strategic pillars. This is not aspiration — this is an operational roadmap with hardware, software, and timeline already in motion.

1. AI Inference at Warp Speed: All local AI workloads running on the GPU workstation's RTX 3080 (or upgraded accelerator). RAG document ingestion fully operational via ChromaDB v2. MCP servers bridging every service to natural-language AI queries.

2. Enterprise-Grade Security (Wazuh & Kali): Wazuh acts as the fleet's internal MXDR sensors — continuously correlating logs across all nodes to detect anomalous behavior before it becomes a breach. Kali operates as a tactical away team for proactive security scanning. This mirrors the enterprise MxDR methodology the Admiral consults on professionally — except we own the SOC.

2.5. RAG → RAG2: Upgrading the Ship's Computer. Standard RAG is like querying the ship's computer with a keyword search — fast, but shallow. Context is found, not understood. RAG2 is a fundamental architectural upgrade to the ship's intelligence layer: multi-hop cognitive retrieval that chains reasoning steps across multiple knowledge sources, high-dimensional embedding vectors mapped directly to the local GPU, and contextual memory that doesn't just find information — it reasons across it. The ship's computer doesn't merely answer questions anymore. It synthesizes. It infers. It anticipates. By implementing RAG2 with denser embeddings and a fully air-gapped local GPU pipeline, the fleet gains an autonomous intelligence core that operates at sovereign speed — zero latency waiting on external APIs, zero data sovereignty risk, and reasoning depth that matches the most capable cloud-hosted systems. This is the upgrade from LCARS 4 to LCARS 7. The difference is not incremental — it is generational.

3. High Availability: Pi4 monitoring Apollo's heartbeat with automated UPS failover via NUT. Proxmox backup jobs writing to genesis-nfs nightly. No single point of failure takes the fleet offline.

4. Zero-Touch Deployment: Cloud-Init/Preseed templates for new containers. OpenClaw autonomously handling routine tasks via Discord. Home Assistant orchestrating the physical environment alongside digital infrastructure.

5. Professional Synergy — The Career Arc: This homelab is not a weekend experiment. It is the direct continuation of a career arc forged in the trenches of enterprise security and cloud services. Starting in network administration — managing physical switches, DNS, DHCP, VLANs, and firewall rules for real organizations — built an instinct for infrastructure that no certification can replicate: you learn how networks actually fail, not how textbooks say they should work. That foundation led to a cloud services pioneer that delivered email security, spam filtering, and Microsoft 365 to thousands of SMB and MSP customers long before "cloud-first" was a board-level mandate. That experience taught what MSPs actually need: reliable, simple, and secure solutions that work at scale without an army of engineers to maintain them. Through a series of strategic acquisitions, that trajectory landed squarely inside one of the world's largest enterprise information management companies — carrying with it the MSP-first DNA, the email security expertise, and the network engineering intuition. Every Wazuh correlation rule built in the homelab reflects an MSP's real-world threat landscape. Every Kali scan mirrors a penetration test a customer once asked about. Every MCP server bridges a gap that MSPs spent years wishing existed. The homelab is professional development disguised as passion. And it has been paying dividends since the first switch was racked.

Translating Lab to Enterprise — Security-First AI Architecture

As these AI capabilities scale from the homelab to enterprise production, a strict Zero Trust security framework governs every layer — built on three core pillars:

• Identity-Aware Access Control: Ephemeral, cryptographically-signed credentials replace static authentication — ensuring every connection is identity-verified, time-limited, and fully auditable. Anomalous access is revoked in milliseconds before lateral movement can occur. Zero Trust is not a feature — it is the architecture itself.

• Real-Time AI Observability: Continuous SIEM-grade monitoring provides real-time visibility into AI agent behavior — correlating logs, flagging anomalies, and enabling autonomous response across the entire inference pipeline.

• Immutable Audit Trail: Tamper-proof logging ensures absolute auditability of all AI actions for enterprise compliance, legal holds, and forensic investigation. Every agent action has a permanent, verifiable record.

Beyond backend infrastructure, a critical capability of the ZeusApollo architecture is the rapid, on-the-fly generation of stunning, responsive dashboards. The data is only as valuable as our ability to make it beautiful, digestible, and interactive for the client.

Recent Dashboard Prototyping: Below are screenshots of custom HTML/JS interactive dashboards generated instantly during customer engagements. These interfaces are designed with bespoke LCARS aesthetics, optional audio feedback (beeps and clicks), and flawless responsiveness.

APOLLO CMD CTR 🔍 CLICK TO ENLARGE

ZEUSAPOLLO INTEL 🔍 CLICK TO ENLARGE

Future IAM & EntraID Visualizations

The true power of this presentation layer unlocks when applied to Enterprise Identity and Access Management (IAM):

• EntraID / Active Directory Topologies: Instantly visualizing complex group nestings, conditional access policies, and orphaned accounts using interactive Vis-Network graphs.
• RBAC (Role-Based Access Control) Matrix: A responsive, color-coded heat map showing exactly which users hold privileged access across tenant boundaries, highlighting anomalies and over-provisioned accounts.
• Customized User-Centric Views: Dynamically morphing the dashboard structure based on what the logged-in user interacts with most. If the CISO logs in, the dashboard immediately bubbles up risk scores; if an IAM engineer or architect logs in, it elevates real-time sign-in logs and lateral movement paths.

Combining the backend zero-trust security of Wazuh with these customized front-end visual layers creates an untouchable enterprise security product.

The Prime Directive — Why Local-First Wins

Before comparing hardware options, the Admiral's standing order must be understood: The Prime Directive of ZeusApollo is local-first, on-premises execution. Every token of inference, every log ingested, every model fine-tuned should run on hardware we own, in a room we control, on a network we defend. Cloud is the fallback — not the default.

This isn't ideology — it's ROI calculus. Local hardware has zero marginal cost per query after the initial investment. There is no "cloud meter ticking" when you're experimenting at 2 AM. There is no data sovereignty question when sensitive client scenarios never leave the LAN. And there is no vendor lock-in when you own the iron. The Prime Directive holds.

CapEx Visual — ROI Break-Even Timeline

Future Arsenal — What Enterprise Practitioners Crave

The homelab is the proving ground. When enterprise testers and MSPs see this architecture in action, they immediately recognize the capability gap in their own environments — and they want it. Here is the operational wish list that creates real conversations at the CXO level:

• BYOai (Bring Your Own AI): Enterprise testers demand the ability to plug their preferred bleeding-edge models — Llama 3, Mistral, Gemma, Phi-4 — into a secure, orchestrated framework and achieve state-of-the-art capability without surrendering sovereign data to public APIs. The homelab demonstrates this is not only possible but production-ready today.

• Rapid LLM Architecture Prototyping — The Irreplaceable Advantage: This is where the homelab creates enterprise value that simply cannot be purchased on a public cloud. The ZeusApollo fleet is the only environment where you can spin up llama3.1:70b at midnight, benchmark it head-to-head against Mistral-7B and Gemma-27B on a live RAG pipeline, swap the entire orchestration architecture to a multi-agent graph, stress-test adversarial prompts against the retrieval layer, document the differential results, and tear it all down — before your morning coffee. No procurement cycle. No cloud bill. No risk to production. No API rate limits pausing the experiment. No data sovereignty exposure. This velocity of experimentation is not available at any price on a public cloud metered usage model. The enterprise value is not merely the capability itself — it is the speed of learning and the freedom to fail fast. When a new model architecture emerges (and they emerge weekly), the ZeusApollo fleet can validate its enterprise applicability in hours, not weeks. This is the competitive moat.

• Orchestrated Security Principles: The absolute mandate to launch AI at the enterprise level safely. Enterprise practitioners require identity-aware access controls, autonomous threat response, cryptographic audit trails, and full observability over AI inference pipelines.

• True Auditability: Beyond basic query logs — this is WORM-compliant cryptographic retention for legal holds. Enterprise teams want immutable proof of exactly why an AI agent pulled specific context and exactly what it returned, ensuring intellectual property containment and regulatory compliance.

• A New Sellable Product: The final frontier is packaging this architecture. Enterprise SaaS providers have a monolithic opportunity to productize the orchestration of local, sovereign, and impregnable AI. Demonstrating that an organization can deploy multi-agent LLM task forces with military-grade Zero Trust security is a massive, untapped market differentiator — and the Admiral already has a working prototype running in his living room.

The Paradigm Shift: From Seats to Consumption

SaaS revenue is undergoing a fundamental inversion. The future relies not on per-seat licenses, but on API consumption. With breakthroughs like MCP (Model Context Protocol), MSPs urgently need bespoke AI solutions to automate complex enterprise tasks. The window for creating this architecture is open right now.

Why I Am The Right Creative Technologist

Zero Trust from Birth: My dad owned a pawn shop. I was raised on "Zero Trust" long before it became an industry buzzword. Security isn't bolted on—it's innate to every architectural decision I make.

The Master Chef Orchestration: I've been cooking since I was 5. Elite model orchestration is exactly like preparing a phenomenal meal: it's not about throwing raw ingredients into a pot. It's about routing specific tasks to the right models at exactly the right moment to create a flawless, automated outcome.

Innovation Born of Pain: Years of consulting with MSPs forged my drive to innovate out of sheer operational friction. Today, armed with this AI fleet, I rapidly prototype solutions in hours—transforming enterprise pain points into demonstrable, production-ready PoCs.

Kinesthetic Leadership: The era of reading whitepapers and theorizing about AI is over. I don't just read about lifting weights; I lift the iron. I've built the network, deployed the models, integrated the SIEM, and orchestrated the API routing. I am not imagining the future of secure enterprise AI. I am already running it in my living room.

Design Philosophy — Rutan × Johnson

My approach to building technology lives at the deliberate intersection of two aerospace legends whose philosophies, taken together, form something greater than either alone: