Download RESUME PDF
View RESUME.md
DevOps-leaning platform engineer building self-hosted AI infrastructure, automation frameworks, and tools that solve real problems I've had.
Currently operating a full local AI stack (Ollama, RAG, agent gateway, coding agents) on consumer hardware with production-grade automation, and looking to bring that same "make infrastructure reliable" mindset to a team that values shipping, not just writing.
| Domain | Evidence |
|---|---|
| DevOps / Platform | Self-hosted Nextcloud, MagicMirror, AI stack — VPN-gated, reverse-proxied, auto-updating with rollback |
| QA / App Testing | Built a multi-layer (Appium + Selenium + API) test framework with Allure + Jenkins CI and a CLI project generator |
| Automation | Cron-orchestrated scraping, AI enrichment, YouTube content pipeline, AI-assisted code generation |
| Local AI / MLOps | GPU LLM serving (6GB VRAM), RAG (ChromaDB), embedding search, agent delegation |
Multi-layer automation testing framework — Appium (Android) + Selenium (Web) + requests (API), unified Page Object Model, Allure + Jenkins CI, Jinja2-powered CLI that scaffolds entire test projects in seconds.
Why it matters: Most testing tools are single-layer. This one runs all three from a shared codebase with a code generator that eliminates boilerplate.
Always-on AI home assistant — wake-word voice (Vosk), local LLM reasoning (Ollama), YOLOv8 vision (custom-trained for my dog + cigar detection), Gmail notifications, persistent memory (Mem0 + ChromaDB), Streamlit dashboard.
Why it matters: Demonstrates end-to-end ML integration — speech → intent → action → memory → response — running reliably as a long-lived daemon.
Multi-board job search platform — scrapes LinkedIn, Indeed, Glassdoor, ZipRecruiter; enriches with local Ollama; tracks applications through a Kanban board; generates tailored interview prep; ships a daily email digest via cron.
Why it matters: Built because existing tools don't talk to each other. Production data flow: scrape → enrich → track → prep → apply — all Dockerized, all local-first.
🐳 ai-stack
Self-hosted AI platform — Ollama (GPU), Odysseus (RAG workspace), Hermes (agent gateway), OpenCode (coding agent), SearXNG, ChromaDB, ntfy, MusicGen, Telegram bots. All Dockerized with NVIDIA passthrough, automated updates, health checks, and rollback on failure.
Why it matters: Real infrastructure work — VPN networking, GPU memory budgeting, automated rollback, config guards after schema migrations. This is the stack that powers the other three projects.
Languages: Python · Bash · YAML · Markdown · a little JS/TS Infra: Docker, Docker Compose, systemd, WireGuard, Tailscale, Nginx Cloud: AWS (S3, EC2 basics), rclone, Backblaze B2 Testing: Pytest, Selenium 4, Appium 2, Allure, Jenkins, Playwright AI/ML: Ollama, ChromaDB, Mem0, Vosk, YOLOv8, OpenAI API Data: SQLite, ChromaDB, RAG, semantic search, BeautifulSoup Ops: Cron orchestration, health monitoring, automated rollback, systemd services
- Nextcloud behind Mullvad VPN — my file cloud + PIM
- MagicMirror² on Raspberry Pi — ambient dashboard
- AI Stack on Ubuntu + RTX A3000 — local LLMs, agents, music generation
- Automated updates — Mon/Thu 9 AM cron pulls new releases, backs up state, rolls back on failure
- Off-site encrypted backups — daily rclone to B2
I don't just write code — I run production. I know what flaky looks like because I've debugged it at 2am on my own stack.
Roles where I can apply: platform engineering, DevOps, SRE, QA automation, test tooling, infrastructure-as-code, or anything involving "make reliable systems that do useful things."
Open to: Remote, hybrid, or relocation (Virginia / Florida preferred). Available immediately.
- Email: maguilar1310@gmail.com
- GitHub: github.com/man1328
- Resume: Download PDF (pinned release)
- This page: github.com/man1328
Last updated: August 2026 This README is auto-curated — if a repo here looks stale, ping me, it's probably private and I can grant read access on request.