How I Accidentally Became a DevOps Engineer
I didn’t study DevOps. I didn’t plan for it. I found out what I actually am when I built a CI pipeline for my genetic algorithm project and felt something click — harder than anything frontend or ML ever gave me.
Turns out I’d been doing DevOps for years without knowing it had a name.
The Pattern Was Always There
Looking back, the signs were embarrassingly obvious:
- Built a Telegram bot so friends could access lectures — that’s infrastructure thinking, not app thinking.
- Loved Linux for years but kept getting pulled toward web and AI because that’s what everyone said to learn.
- Built and sold Notion templates at 15–16 years old — and never used them myself. I liked building the system, not using it.
- Always building things for the sake of building them perfectly, not for personal use.
The Obsidian and Notion obsession isn’t separate from DevOps — it’s the same brain applied to a different domain. Organizing systems, automating workflows, making everything connect cleanly.
What I’d Already Built (Without Studying Any of This)
The home server — Two old laptops, an HP and a Dell, split by exposure, one facing the internet and one that never did. Pi-hole, Gitea, and Samba ran bare metal on the LAN box; the public box ran a small container stack behind nginx. Hardened with SSH keys, a custom port, Fail2ban, UFW, and WireGuard running natively on the host rather than in a container. Solved Syria’s CDN blocks by downloading packages on my main laptop and SCP’ing them to the server.
Terminal Toolkit — A unified tool command launcher with a shared library system (lib/colors.sh, lib/ui.sh, lib/utils.sh, lib/config.sh), config overrides, and scripts for health checks, Docker cleanup, Ollama management, git workflows, backups. Built on top of Oh My Zsh and a lot of bash tricks. This isn’t a script folder — it’s a designed system with a shared library, a unified interface, and documentation. That’s software architecture.
pdfpeek — Published on PyPI. Built in 2 days. 10-stage PDF extraction pipeline with confidence scoring. Tiered installs, CLI + Python API. Has a roadmap, comparison table, known limitations. Shipped like an actual product.
What I Actually Am
Not: Frontend developer. ML researcher.
Actually: A systems engineer and tooling builder who thinks in infrastructure, automation, and design systems.
The moment I stopped trying to be what everyone else was studying and looked at what I kept building on my own time, the answer was obvious.
See Also
- Self-Hosting on Two Old Laptops — the project that made it click
- The Productivity Trap
- Running LLMs Locally with Ollama
- WSL for Development