Patrick Moore

AI Systems Architect & Full-Stack Developer

I build production AI systems, automation pipelines, and the tools that ship them. 22+ years writing code. Currently building LLM integration systems, semantic routing engines, and tools that make AI actually useful — not just impressive.

Junction City, KS · Remote-ready · US Citizen

kit@holdtheline.tech GitHub
🚀

By the Numbers

100+
Automation tools built
8
Live production apps
22
Years writing code
120+
Routes in main PWA
Digital earth network visualization

Experience

AI Systems Architect & Developer

K Systems (Independent)
2024 – Present

Designed and built a semantic routing engine that classifies user intent and routes queries to optimized response pipelines — achieving 80% template deflection and significant cost reduction vs. raw API calls. Built 100+ MCP (Model Context Protocol) automation tools. Shipped 8 production web applications on Vercel with Supabase backends. Developed customer service AI platform with BYOK model support and multi-tier pricing.

Python TypeScript SvelteKit Claude API MCP Servers RAG

AI Implementation Consultant

Generali
2024

Contributed to an AI-to-PDF RAG system implementation for document processing workflows. Assisted with pipeline architecture, embedding strategy, and retrieval optimization for insurance document ingestion and query.

RAG PDF Processing LLM Integration Enterprise AI

Technical Support Engineer (Tier 2/3)

Spectrum / Time Warner Cable
10 years

Hired as Tier 1, quickly handling Tier 2 escalations and Tier 3 issues independently. Deep DOCSIS expertise — modem provisioning, signal analysis, plant troubleshooting, firmware diagnostics. Managed complex escalation workflows across support tiers. Built internal automation tools to streamline repetitive diagnostic procedures.

DOCSIS Telecommunications Tier 2/3 Support Escalation Management

Software Developer & Automation Specialist

22 years

Started with Lua scripting in 2004. Built game mods, automation tools, MUD systems, and install wizards across two decades. Transitioned to professional full-stack development with Python, TypeScript, and modern AI toolchains. Deep experience building deployment and installation tooling for complex software packages.

Lua Python Install Wizards Game Modding Automation
Built & Deployed

AI Customer Service Agents

A production customer service AI platform that works for any industry. Not a demo — live, handling real queries, with real cost savings.

How It Works

Incoming customer queries hit a semantic routing layer that classifies intent and routes to the right response pipeline. 80% of queries are handled by optimized templates without ever touching an LLM API call — cutting costs by up to 48x vs. naive GPT/Claude implementations.

The remaining 20% escalate through a tiered AI stack: small model for simple cases, large model for complex ones, human handoff when needed. Full audit trail on every interaction.

Industry-Agnostic

Plug in any knowledge base. Insurance, legal, telecom, retail, SaaS — the routing layer adapts.

Multi-LLM / BYOK

Works with Claude, GPT, Gemini, or local models. Bring your own keys. No vendor lock-in.

Escalation Logic

Template → small model → large model → human. Each tier only fires when needed.

Audit & Compliance

Every interaction logged. AES-256-GCM field-level encryption. Built for regulated industries.

SvelteKit Python Supabase Claude API MCP Servers 100+ Connectors Vercel
Server room with blue lighting

What Else I Build

RAG & Document Processing Pipelines

End-to-end retrieval-augmented generation systems. PDF ingestion, embedding generation, semantic search, and contextual response generation. Built for enterprise document volumes with audit logging and field-level encryption.

RAG Embeddings AES-256-GCM PostgreSQL

MCP Automation Tooling

100+ Model Context Protocol servers that give AI direct access to business tools. Custom connectors for any REST API. Microsoft 365 integration, calendar intelligence, document automation, notification systems. If your tool doesn't have a connector, I build one.

MCP Claude REST API M365

Install Wizards & Deployment Tooling

Step-by-step installation and configuration wizards for complex software packages. Environment detection, dependency resolution, configuration validation, and guided setup flows. Making deployment painless for end users.

Python PowerShell Cross-platform

Live Products

Shipped, deployed, running in production right now.

Golden light through cracks

Held

ADHD management PWA. Phase-aware scheduling, skew detection, offline-first. 120+ routes, 60+ data stores.

LIVE
Matrix-style code screen

Saga.Logo.AI

Dual-persona AI writing coach with split-panel editor. BYOK multi-provider support.

LIVE
Technology network connections

HoldTheLine

Customer service AI platform. Tiered pricing, knowledge base routing, template deflection.

LIVE
Earth from space - NASA

Technical Skills

Languages

  • Python
  • TypeScript
  • JavaScript
  • Lua
  • SQL
  • PowerShell

AI / ML

  • RAG Pipelines
  • LLM Integration
  • Prompt Engineering
  • MCP Servers
  • Embeddings
  • Local Models
  • Claude API
  • Multi-provider

Frameworks

  • SvelteKit
  • Tailwind CSS
  • Vite
  • Node.js

Infrastructure

  • Vercel
  • Supabase
  • PostgreSQL
  • Cloudflare
  • Git / GitHub
  • PWA
  • CI/CD

How I Work

⚡

Ship First, Iterate Fast

I'd rather have working software in your hands today than a perfect spec next month. Deploy early, refine from real usage.

📈

Cost-Conscious AI

Raw API calls are expensive. Semantic routing and template deflection handle 80% of queries without the LLM — 48x cost savings.

🔌

I Build the Connectors

Most AI consultants configure what exists. I build what doesn't. Need AI to talk to your tools? Working connector in hours, not weeks.

Let's Build Something

Looking for someone who builds AI systems that actually work? Available for consulting, contract work, and full-time opportunities.

Get in Touch View GitHub