The shop floor
I learned software where mistakes cost money by the minute. Online shops, secure payment gateways, product feeds for Skroutz and Google Shopping.
And an ERP that had to agree with the website about what was in stock: every product, every order, every day.
Let the machines do it
Some work is the same click a thousand times. I wrote Python tools that talked to Slack and Google Drive, scraped what needed scraping, and turned spreadsheets into structured WordPress pages.
Custom plugins and Gutenberg blocks did the rest, so the people running the shop could stop copying and pasting.
The lab
Then the route bent toward research. At a spin-off of Aristotle University I moved from storefronts to systems that serve hospitals and scientists.
Django and Angular on the surface. PostgreSQL and Docker underneath. Partners in several countries.
The bug hunt
A bug lives in the client, the API, the container network, or a partner’s server in another country. Finding which one is the job.
I read unfamiliar code until it makes sense, follow the request layer by layer, and fix it where it actually breaks.
Campus Safety: finding someone indoors
GPS stops at the door. For an emergency-response system at Aristotle University, I worked on indoor positioning: knowing where someone is inside a building when it matters.
One system, from the phone in a pocket to the model that answers where.
Ten million records, one shape
Hospitals and institutes across Europe each store patient data their own way. My part: containerised pipelines that take each partner’s data, reject what is broken, anonymise the rest and load it into one shared research database, under GDPR.
Scattered on the left, research-ready on the right. That is the whole job, ten million times.
The thesis
I finished my degree on the Travelling Salesman Problem: given a set of places, find the shortest tour through all of them. I benchmarked neural networks, heuristics and classic algorithms against each other.
It rarely finds perfection. It finds good, quickly, and that is most of engineering.
The toolbox
Six groups of tools, picked up one problem at a time.
- Languages
- Python, TypeScript, JavaScript, PHP, SQL
- Backend
- Django, Flask, REST, OAuth, WordPress
- Frontend
- Angular, Ionic, React, HTML, CSS
- Data
- PostgreSQL, MySQL, ETL, Validation
- Infra
- Docker, Compose, Linux, Deployment
- ML
- PyTorch, LightGBM, XGBoost, CatBoost, OR-Tools
The next stop is open.
If you have a system that crosses too many layers, or data that needs to be trusted, I would like to hear about it.
Maps: Natural Earth · regions: geoBoundaries (CC BY 4.0)