In 2021, my capstone project at Northern Technical University was a CNN-based traffic-sign recognition system built in Python: data preprocessing, model training, evaluation, the full pipeline. I graduated ranked 9th in my class with a GPA of 2.81 out of 4.0, a solid but unremarkable result on paper. Five years later, I am running production AI automations for real businesses and helped ship a live e-commerce site built largely with an AI coding agent. The line between those two points is not a straight one, and I think the messy version is more useful than the polished one.
What the capstone actually gave me
Building that traffic-sign recognition system meant sitting with the actual mechanics of how a neural network learns: how data quality shapes model performance, how training and evaluation are two different problems, and how a model that looks accurate on paper can still fail in ways that matter. That foundation is the reason I was never intimidated by AI models later on, when ChatGPT, Claude, and Gemini became mainstream tools. I had already spent months elbow-deep in a simpler version of the same underlying idea.
The unglamorous years in between
After graduating, I did not walk into an AI engineering role. I did freelance software testing on Upwork, testing a live survey application end to end. I built Telegram bots for fun and for small clients, contact-splitter bots, image processing tools, notification automations, using AI-assisted development to accelerate the work. None of that was glamorous, and none of it looked like a straight line toward "AI automation developer" as a job title. It was just consistent tinkering with real, if small, problems.
Where the marketing and accounting work fits in
At the same time, I was running social media marketing and, later, accounts and inventory management at Help Tech Co. Ltd. Those roles are not unrelated detours from the automation work; they are where the automation work actually came from. I started using n8n specifically to stop doing the same weekly reporting by hand. That practical need, not a career plan, is what turned into the automation practice I run today. You can see the full arc across roles in Experience.
From automation scripts to production AI agents
The current chapter is production n8n workflows: an Invoice Fraud Firewall, an AI Lead Intelligence and Auto-Response system, a Multi-Stage Invoice Approval Pipeline, all integrating OpenAI, Claude, and Gemini into real business processes with proper error handling, retry logic, and audit logging. Alongside that, DAD LINK is a live e-commerce store built largely with Claude Code, proof that AI-assisted development can ship something a real business actually runs on, not just a demo.
None of this happened because I planned a career path toward "AI automation developer." It happened because I kept saying yes to the next unfamiliar problem, and eventually the problems all pointed the same direction.
What I would tell someone starting where I started
If you are a computer engineering or computer science graduate wondering whether your degree "counts" without a big-name internship or a straight-line career plan, my honest answer is that the degree gets you the foundation, but the actual skill gets built in the unglamorous years of freelance testing, small bots nobody remembers, and jobs that were not officially about AI at all. If you want to talk about how to apply that kind of path to your own automation or development needs, or just compare notes, reach out.