Solutions Engineer · AI & DevOps Infrastructure

Solutions Engineer & AI Systems Architect

I run technical discovery, build and deliver POCs, and translate complex AI and DevOps infrastructure into clear outcomes for customers and stakeholders.

Discovery Architecture POC Outcomes

Discover. Architect. Deliver.

This portfolio showcases customer-relevant systems work: architectures, technical tradeoffs, reliability practices, and POC-grade implementations with measurable results. It captures the work that doesn't fit on a standard CV but makes the difference in a technical discovery conversation or a live proof-of-concept.

My career started in technical support — five years as the escalation point for customers with the most complex problems. That instinct to translate technical complexity into clear outcomes never left. Every role since has required making difficult systems understandable: to customers, to stakeholders, to product teams.

I lead architecture, governance, and verification for high-stakes systems — using AI as an accelerator, with human-owned design reviews, test strategy, and clear accountability. The goal is always deployable, defensible outcomes.

Flagship Projects

FireCalc Pro

Life-Safety Fire Alarm Load Calculation Tool
In Progress
Users: fire protection engineers & estimators Outcome: faster compliant submittals, fewer rejections, traceable revisions

A life-safety engineering tool for fire alarm system load calculations. Architected with a domain-driven schema capturing device-level electrical specifications (NAC circuits, SLC, standby vs. alarm current draw) required for code compliance. Features an immutable CalcSnapshot pattern for tamper-evident historical records and a state-machine project lifecycle (Draft → Quoted → Approved → Locked) enforcing strict engineering accountability.

Includes a full revision audit trail capturing actor, action, and JSON diffs per change — establishing a complete chain of custody essential for life-safety engineering and financial accountability in AI-assisted code.

Python Automated E2E Verification Life-Safety / Fire Code Human-led Verification Isolated Load Math

Zoho CRM Consulting

End-to-End Client CRM Delivery
Complete

Ran a full client engagement from discovery through delivery. Conducted requirements sessions with sales ops, leadership, and end users; configured custom module schemas and lead-lifecycle automation; and delivered a production CRM from zero baseline. Conducted a formal Buy vs. Build evaluation comparing Zoho against the open-source Moqui ERP framework, producing a documented TCO and risk analysis.

Result: client adopted the platform recommendation; full CRM and training handoff delivered; reduced manual lead handling. Complete pre-sales and post-sales motion executed independently.

Zoho CRM Requirements Discovery TCO Analysis Client Delivery Moqui ERP

MeshBrain Cluster

8-Node Distributed AI Coordination Platform
Active

A heterogeneous 8-node compute fabric spanning Apple M2 (MLX-LM), RTX 3070 / GTX 1080 (Ollama), Dual Xeon E5-2609, and RPi 5 / RPi Zero 2W edge nodes. A custom-built mesh-router (OpenAI-compatible FastAPI service, k3s) provides unified API routing with JWT lease tokens, Postgres-backed lane management, model viability checks, and dualboot mutual exclusion across GPU workers. Built a purpose-fit router after hitting reliability limits with off-the-shelf gateways.

Coordination layer uses PostgreSQL + pgvector for semantic memory and embedding-based recall, and Redis pub/sub for real-time inter-agent messaging. Engineered a persistent agent-to-agent memory handoff protocol enabling multiple LLMs to collaborate on a single codebase across sessions. Manages 22TB LVM storage and a k3s-orchestrated service stack. Why it matters: predictable inference routing, safer multi-model workflows, and operational controls that prevent runaway cost and latency.

mesh-router PostgreSQL/pgvector Redis pub/sub k3s Python MLX-LM Ollama

BCWPayroll Swarm Audit

Multi-Agent Compliance System
Complete

An agent-assisted audit system using 5 parallel AI instances in a voting architecture to surface FLSA risk areas in complex payroll logic. Each agent independently analyzes overtime calculation and federal withholding logic; results are reconciled to identify high-confidence risk patterns.

Deliverable: structured audit report, reproducible test harness, and prioritized remediation checklist — validated by stakeholder review. A case study in applying AI to high-stakes compliance work with human-owned verification strategy.

Python Multi-Agent Voting FLSA Compliance Parallel AI Inference AI Governance

mesh-router

Custom OpenAI-Compatible LLM Router
Active

A homegrown replacement for LiteLLM, purpose-built to solve API token leasing and routing reliability issues across a heterogeneous GPU/CPU/MLX cluster. Deployed in k3s as an OpenAI-compatible FastAPI service with a dedicated Postgres database.

Features JWT-signed lease tokens as the single source of truth for in-flight LLM requests, Postgres-backed lane management with model viability and swap-history tracking, readiness probes that never trigger model loads, and dualboot mutual exclusion to safely arbitrate between shared GPU workers.

PythonFastAPIPostgreSQLJWT / Lease Tokensk3sOpenAI-Compatible API

Engineering & Home Lab

Image: ZeroG Build / CAD Models

ZeroG Mercury One.1 & CAD Design

Active

CoreXY conversion of a Creality Ender 5 Pro utilizing a split-MCU architecture (Two BTT SKR Picos + RPi 4 SBC) to drive 7 steppers. Engineered a 750W 110VAC heated bed via SSRs.

CAD: Designed custom printer components and functional outdoor infrastructure, including fence post toppers with internal routing for future electrical integration.

Klipper/RPi4 SKR Pico AC Mains Power CAD/Fusion360
Image: Custom 3D Printed Doorbell Unit

Integrated Security & IoT Control

Complete

Smart Doorbell: Built a custom 2-way audio video doorbell (Pi Zero 2W) with camera streaming via go2rtc RTSP into Frigate NVR for AI-based object detection, triggering push notifications through MQTT to Home Assistant.

Security Panel Hack: Repurposed a dead alarm panel's hardwired contact loops by soldering a separate Pi Zero directly to the existing sensor wiring. Powered via Waveshare PoE hat (pogo pins) over Cat 5e for real-time door/window announcements.

Garage Controller: RPi-based garage door controller with sensor state monitoring and remote operation.

Frigate NVR PoE / Cat 5e GPIO / Relay Logic go2rtc / RTSP Soldering
Image: Server Rack / Home Assistant

Enterprise Home Lab & Automation Stack

Active

Dual HP Proliant Gen 8 1U servers (4x 8TB drives each) with mixed RAID configurations (RAID 0 for media performance, RAID 5 for backup redundancy). 22TB LVM storage fabric. Unified fragmented smart home brands into Home Assistant (RPi 5 w/ NVMe), engineering logic-based automations for sun-activity lighting and power-saving "kill switches."

k3s-orchestrated service stack: Frigate NVR (AI object detection), Jellyfin media streaming, WireGuard VPN (10 peers), Uptime Kuma monitoring, Watchtower auto-updates, Nginx Proxy Manager, Borg automated backups, n8n workflow automation, mesh-router (custom OpenAI-compatible LLM router, replaced LiteLLM), and game servers.

HP Proliant RAID 0 / RAID 5 22TB LVM Home Assistant k3s WireGuard n8n

Infrastructure Stack

MeshBrain Compute Fabric

  • Apple M2 (MLX-LM inference)
  • Ubuntu node — RTX 3070 (Ollama)
  • Windows node — GTX 1080 (Ollama)
  • Dual Xeon E5-2609 server
  • RPi 5 (Packserv) — k3s control plane + mesh-router host
  • RPi Zero 2W — edge nodes
  • 8 nodes total, 22TB LVM storage

Orchestration Layer

  • mesh-router — custom OpenAI-compatible router, JWT lease tokens, lane viability
  • MeshBench — multi-host model benchmark harness with live dashboard
  • PostgreSQL + pgvector — semantic memory
  • Redis pub/sub — inter-agent messaging
  • k3s — container orchestration
  • n8n — workflow automation
  • Nginx Proxy Manager — reverse proxy

AI Toolchain

  • Claude Code — primary coding agent
  • Gemini CLI — planning & research
  • Codex — targeted code generation
  • Local LLMs — air-gapped sensitive work
  • MISSION_BRIEF.md protocol — agent governance
  • Watchdog — session health monitoring
  • MCP Servers — tool augmentation

Core Technologies

Python PostgreSQL pgvector Redis Docker mesh-router k3s MetalLB Java Spring Boot AWS ECS .NET / C# Vue.js FastAPI Klipper Home Assistant WireGuard Nginx GitHub Actions Cloudflare Pages

Professional History

Independent Systems Architect & AI Infrastructure Engineer

Freelance / Self-Directed
Jul 2025 - Present

Designing and operating distributed AI infrastructure while directing an agentic workforce to build and verify mission-critical software. Architecture and governance are 100% human-authored; AI agents execute within strictly defined constraints.

  • Delivered end-to-end CRM engagement: requirements discovery, solution design, TCO analysis, technical configuration, and client training handoff
  • Translate complex AI architectures and technical tradeoffs into clear business strategy for non-technical stakeholders
  • Architected MeshBrain: 8-node compute cluster with custom OpenAI-compatible LLM router, PostgreSQL/pgvector semantic memory, and Redis inter-agent messaging — available as live demo environment
  • Led agent-assisted FLSA compliance review: structured audit report and remediation checklist delivered to stakeholders
  • Built FireCalc Pro: life-safety fire alarm load calculation tool with tamper-evident audit trail and state-machine project lifecycle

Software Engineer

Taurus LLC
Dec 2024 - Jul 2025

Backend modernization and cloud infrastructure. Updated microservices in Python and PHP, managed AWS ECS deployments, and implemented CI/CD pipelines via Github Actions.

AWS Python Docker

Software Engineer

Derivco
Aug 2023 - Sep 2024

Distributed systems development for high-volume transaction platforms. Led initiatives to upgrade internal tools from .NET 4 to .NET 8 and modernized legacy Silverlight apps to Vue.js.

Software Engineer

ReliaQuest
Jul 2021 - May 2023

Enterprise security platform development. Reduced data-ingestion latency 2 min → 7 sec on customer-facing pipeline, directly preventing SLA breaches for enterprise security clients. Served as on-call incident commander: led cross-team coordination under production pressure with real-time stakeholder communication via Datadog/Splunk.

Software Engineer

Xactware (Verisk)
2009 - 2021

Software Engineer (2018 - 2021): Built microservices for roofing data integration and automated test frameworks using Selenium/Protractor.

QA Automation Engineer (2014 - 2018): Advanced from manual testing to writing automated test suites for web applications.

Technical Support Tier 1 & 2 (2009 - 2014): Started in frontline support, rapidly advanced to Tier 2 resolving complex technical issues for adjusters.

Systems & Low Voltage Technician

Why'rd
2005 - 2009

Physical Infrastructure: Installed commercial and residential systems including fire/burglar alarms, CCTV (RG59), and card access.

System Integration: Factory trained (Control4 HQ) in home automation programming. Managed structured cabling (Cat 5e, Coax RG6) and control loops (18/2, 22/4).

Technical & Sales Support

Sento
2004 - 2005

Provided high-speed internet support for hotel guests (STSN) and later transitioned to sales support for Overstock.com.

Bachelor of Science, Computer Engineering

Utah Valley University (2013 - 2019)

Repositories

Private Repositories

mesh-computer

Python/React — Autonomous multi-step job orchestrator with LLM planning and tool execution

firewire-firecalc-pro

Python/FastAPI — Life-safety fire alarm load calculation tool

BCWPayroll

Python — Payroll system subject of the BCWSwarm FLSA compliance audit

packdoorbell

Python/RPi — DIY smart doorbell with RTSP streaming and Frigate AI object detection

packgarage

Python/RPi — Garage door controller and sensor dashboard