AI Infrastructure & Automation
I build the infrastructure that makes AI work.
Multi-agent orchestration, autonomous pipelines, and production tools that keep running whether I'm at the keyboard or not.
Work
Projects
Everything here is in production. Real infrastructure, real users, real revenue potential.
Blu Print Engine — Autonomous Multi-Agent AI Platform
Production4 Agents. 21 Jobs. Zero Human Intervention.
Self-hosted AI operations platform running 24/7 on a Linux VPS. 4 specialized agents (Scout, Sentinel, Architect, Nimbus) execute 21 scheduled jobs autonomously — market research, competitive intelligence, VPS health monitoring, strategic analysis, daily briefings, and accountability tracking. All reporting flows through Telegram. Architecture mirrors patterns used in enterprise AI systems: isolated agent responsibilities with defined trust boundaries, structured handoff protocols, and centralized orchestration with decentralized execution.
Anthropic Claude, Google Gemini, and Perplexity Sonar Pro with automatic fallback chains. Each model selected based on task fit — Claude for reasoning-heavy analysis, Gemini for high-throughput summarization, Perplexity for real-time web research.
/spark [idea] triggers a full pipeline: AI-powered market research, GO/MAYBE/NO-GO validation, tech stack planning, Claude Code builds in sandboxed tmux, auto-test, Vercel deploy, and sales packaging. Cost: $0.03 per research cycle. Pipeline exists because manual idea validation was a 2-day bottleneck — automating it removed the feedback loop entirely.
Telegram-triggered Claude Code sessions with protected zone hashing, file integrity monitoring, concurrent session management, and automatic queue processing.
Atomic JSON writes, Syncthing cross-device sync, cron scheduling, systemd services, UFW firewall, watchdog monitoring, logrotate, and automatic failure recovery.
HMAC-signed stateless auth, trust boundaries between agents, guardrails for autonomous operation, SSH hardening, localhost-bound services. Every agent has defined permissions — no agent can escalate beyond its scope.
ACES Aid — AI-Powered Job Aid for Enterprise Technical Support
Production162 Workflows. 1,500+ Keywords. Zero Dependencies.
Solo-built operational platform encoding institutional knowledge into software. Used on live calls daily in a high-volume enterprise support operation. Identified workflow gaps — agents were switching between 8+ browser tabs and losing call context — then built structured troubleshooting tools with specification-driven development to eliminate that overhead. Presented to and adopted by operations leadership.
Single HTML file, zero external dependencies, runs offline on any browser. Deliberate constraint: the production environment had restricted software installation policies and unreliable network connectivity. Zero IT approval needed, zero failure modes from CDN outages.
23 categories with dynamic field population. Each flow encodes the decision tree that experienced agents carry in their heads — making tier-1 agents operate at tier-2 speed from day one.
Fuse.js-powered search across 1,500+ indexed keywords and 90 aliases. Agents describe problems in their own words and get matched to the right workflow. Alias mapping bridges the gap between customer language and internal documentation.
Auto-populates call documentation templates — case notes, escalation forms, incident reports, and account records. Eliminated copy-paste errors and cut documentation time by standardizing output format.
Phone directory, callback scheduler, EOD report export, multi-timezone clocks, device tracking, equipment lookup, edge case detection. Every tool exists because an agent had to leave the platform — each addition removed one more tab.
Device configuration guides, network diagnostics, connectivity settings, signal metrics, error codes, and service policies. Maintained as a living reference updated weekly based on new issue patterns.
Content Security Policy headers, local-first data storage, zero external network calls.
SoloInvoice AI
ShippedAI-Powered Invoice Generator
First product shipped through the Spark Pipeline — idea to production in under 2 hours at $0.03 total research cost. Built to validate the pipeline itself: if it could produce a deployable product with real utility, the pipeline works. Natural language invoice creation, line item editor, logo upload, template management, and PDF export.
RAG Agent
Open SourceDocument Intelligence System
Upload documents, ask questions in natural language, get answers with source citations. Built to demonstrate context architecture over naive vector embeddings. Token-aware chunking preserves semantic boundaries instead of splitting mid-paragraph. Multi-turn conversation memory, source citation evaluation, and retrieval quality scoring that surfaces when the system lacks sufficient context — it says "I don't know" instead of hallucinating. Dockerized.
SQL Agent
Open SourceNatural Language Database Interface
Query any PostgreSQL database using plain English. Schema introspection reads database structure before generating queries — no hallucinated table names. SQL injection prevention and read-only execution enforced at the connection level, not just the prompt level — defense in depth against prompt injection. LangChain agent with 12 unit tests. Dockerized.
Email Agent
Open SourceAI Email Triage & Automation
Autonomous email management with categorization, priority scoring, thread summarization, and smart reply drafting. Pluggable provider architecture means swapping email backends doesn't require rewriting business logic. Priority scoring uses sender history, thread length, and content signals — not just keyword matching. 12 unit tests. Dockerized.
Background
About
Founder and AI Engineer leading production autonomous systems. I designed and operate a multi-agent AI platform with 21+ autonomous jobs across four specialized agents — handling market research, competitive intelligence, infrastructure monitoring, and product deployment with zero human intervention.
On top of it, I built the Spark Pipeline: a single Telegram command takes a raw idea through AI-powered market research, validation, code generation, auto-testing, Vercel deployment, and sales packaging. First product shipped in under 2 hours at $0.03 per research cycle.
Separately, I solo-built an AI-powered job aid for a high-volume enterprise support operation — 162 troubleshooting flows, fuzzy search across 1,500+ keywords, template generation. Used on live calls daily by 40+ agents. Adopted by operations leadership.
My background spans 14 years — from running my own business for 12 years (leveraging Python automation and Docker) to enterprise IT, and now AI engineering. I've seen how real users break software and how manual processes drain resources. I build tools people actually use.
Seeking to apply deep AI engineering and operations expertise to deliver reliable, real-world AI products.
Production-first philosophy: I don't build demos. Every project in this portfolio runs in production or shipped to real users. Production systems need failure handling, security boundaries, cost controls, and monitoring that demos never require. I'd rather ship one system that survives its first week than ten prototypes that work in a notebook.
Failure-mode thinking: I design for what goes wrong, not just what goes right. Fallback chains when APIs go down. Watchdog monitoring when processes hang. Atomic writes when concurrent agents collide. Trust boundaries when autonomous systems need guardrails. The interesting engineering isn't making AI work — it's making AI work reliably, at 3am, with no one watching.
Tech Stack
Process
How I Work
Architect
Design systems that run autonomously, not tools that need babysitting. Every project starts with failure modes: what breaks, what recovers, what alerts.
Orchestrate
Deploy AI agents across models and APIs to execute at scale. Route to the right model for the right task. Build fallback chains so nothing fails silently.
Ship
Idea to deployed product in hours, not months. Live URLs, real users, real feedback. The Spark Pipeline exists because shipping speed is a competitive advantage.
One founder. The right architecture. A system that runs 24/7 and ships products while most teams are still in planning. The live infrastructure speaks for itself.
Capabilities
The 7 AI Skills That Matter in 2026
These are the skills driving AI hiring demand and how I demonstrate each one in production.
Specification Precision
Structured prompt specifications across multi-model orchestration; prompt engineering for 162 operational workflows. Example: ACES Aid's fuzzy search maps 90 aliases to canonical workflow names because customers and documentation use different language for the same problem.
Evaluation & Quality Judgment
Output quality validation across RAG, SQL, and Email agents; 24 combined unit tests; retrieval quality scoring. Example: RAG Agent's retrieval scoring rejects low-confidence answers rather than presenting them — the system says "insufficient context" instead of guessing.
Multi-Agent Decomposition
4 specialized agents decomposing business objectives into 21 scheduled tasks; subagent orchestration. Example: Scout researches, Architect analyzes, Sentinel monitors, Nimbus orchestrates — no agent touches another agent's domain, preventing cascading failures.
Failure Pattern Recognition
Structured fault diagnosis across API, model, and infrastructure layers; automatic fallback chains; watchdog monitoring. Example: When an API returns overloaded errors, Blu Print Engine automatically routes to alternate models for non-critical tasks and queues priority work for retry — zero human intervention.
Trust & Security Design
HMAC-signed auth, RBAC, PCI DSS compliance, NIST frameworks, guardrails for autonomous operation. Example: Blu Print Engine's build pipeline uses protected zone hashing — each agent's workspace is integrity-checked before and after execution, preventing cross-agent corruption.
Context Architecture
RAG over vector embeddings, token-aware chunking, model-specific context window strategies, retrieval optimization. Example: RAG Agent chunks documents at semantic boundaries (paragraph and section breaks) rather than fixed token counts, preserving meaning that fixed-size chunking destroys.
Cost & Token Economics
$0.03 per research cycle, deliberate token optimization, multi-model cost arbitrage across Claude/Gemini/Perplexity. Example: Spark Pipeline routes market research to Perplexity (web-optimized), analysis to Gemini (high throughput), and final synthesis to Claude (highest reasoning quality) — each model handles what it's best at.
Resume
Experience
Founder & AI Engineer
Feb 2025 – PresentBlu Print Solutions, Phoenix, AZ
- Architected and operate a production multi-agent AI platform running 21+ autonomous cron jobs on a self-hosted Linux VPS with 24/7 uptime and automatic failure recovery
- Engineered multi-model orchestration across Anthropic Claude, Google Gemini, and Perplexity Sonar Pro with automatic fallback chains and cost-optimized routing
- Built the Spark Pipeline: single Telegram command triggers AI market research, validation, code generation, auto-testing, Vercel deployment, and sales packaging; first product shipped in under 2 hours at $0.03 per research cycle
- Designed trust boundaries and security architecture using HMAC-signed stateless auth, guardrails for autonomous agents, SSH hardening, and UFW firewall, preventing unauthorized access and ensuring compliance
- Deployed 4 specialized AI agents (Scout, Sentinel, Architect, Nimbus) handling market research, competitive intelligence, infrastructure monitoring, and strategic analysis with zero human intervention
Central Technical Expert
Aug 2024 – PresentValor Global, Phoenix, AZ
- Solo-built an AI-powered job aid for a high-volume enterprise support operation with 162 troubleshooting flows, fuzzy search across 1,500+ keywords, and template generation — adopted by operations leadership for live support workflows
- Built a production call documentation tool used daily by 40+ agents in an enterprise technical support environment with structured data capture and timezone-aware timestamps
- Served as night-shift Central Technical Expert with escalation authority for complex network, device, and account issues, using Zendesk to resolve escalated tickets quickly
- Coached non-technical agents on AI-assisted tooling integration, increasing team-wide tool adoption across live call workflows
- Established standardized procedures aligned with ITIL best practices, reducing tribal-knowledge dependency and improving first-call resolution
Service Desk Support Analyst
Feb 2023 – Apr 2024Massage Envy (Corporate), Scottsdale, AZ
- Managed provisioning and deprovisioning workflows across 1,000+ franchise locations using ServiceNow
- Led network and security audits, including firmware upgrades across 1,000+ devices while maintaining 99.8% uptime across franchise IT infrastructure
- Authored standardized SOPs adopted company-wide, reducing support-ticket backlog and improving team onboarding
- Provided Level 1-2 support for POS systems, network infrastructure, and franchise IT operations via ServiceNow, resolving tickets within SLA
IT Support Specialist II
May 2022 – Jan 2023Alorica, Remote
- Delivered Level 2 technical support for Dell Technologies enterprise clients across desktop, Windows/macOS, mobile, and AV equipment using Zendesk, consistently meeting SLA targets
- Authored 70+ knowledge-base articles covering configuration procedures, testing protocols, and systematic troubleshooting methodology, increasing first-call resolution rates
- Improved resolution time through root-cause analysis and structured documentation across device, software, and connectivity categories
Independent Business Owner
Oct 2010 – Mar 2022Self-Employed, Goodyear, AZ
- Ran an independent service business for 12 years — full P&L, client acquisition, vendor negotiations, and regulatory compliance; leveraged Python automation on Linux and Docker containers to streamline invoicing and service delivery
- Managed end-to-end operations including scheduling, inventory, customer retention, and financial reporting using Excel and a custom Python dashboard
Technical Skills
AI & Automation
Multi-Agent System Design · LLM Orchestration · Autonomous Pipeline Development · Prompt Engineering · Model Context Protocol (MCP) · RAG · Agentic Workflows · Structured Output · Tool Use · Failure Pattern Recognition · LangChain
Development
Python · JavaScript · TypeScript · React · Next.js · Node.js · HTML/CSS · Bash · REST APIs · Streamlit · Tailwind CSS
Cloud & Infrastructure
Amazon Bedrock · Google Cloud Vertex AI · Docker · Linux (Debian) · Vercel · Firebase · Git · GitHub Actions · ChromaDB · Redis · Qdrant · FAISS
Security & Compliance
PCI DSS · NIST SP 800-53 · NIST CSF · SSH Hardening · HMAC Auth · UFW · File Integrity Monitoring
Operations & Service Management
ServiceNow · Zendesk · ITIL · SOP Development · Technical Documentation · Vector Embeddings · PostgreSQL
Certifications
Anthropic Academy (2026)
- ✓Building with the Claude API
- ✓Introduction to Agent Skills
- ✓Claude Code in Action
- ✓MCP: Advanced Topics
- ✓Introduction to Subagents
- ✓Claude with Amazon Bedrock
- ✓AI Capabilities and Limitations
- ✓AI Fluency: Framework & Foundations
- ✓Claude with Google Cloud Vertex AI
Education
Grand Canyon University
BTech, Cyber/Computer Forensics & Counterterrorism
PDF available on request
Get in Touch
Contact
Open to AI Platform Engineering, Infrastructure, and Automation roles. On-site, remote, or hybrid — open to travel. If you're building something autonomous, let's talk.
Glendale, AZ · On-site, remote, hybrid · Open to travel