YPSILON

This is a crazy story, buckle up. I really enjoy walking across the Barcelona Port Vell marina: open space, palm trees, the water just a few meters away… a perfect place to unwind, reflect, and maybe sip a good cocktail.

It was late 2018, and I was walking there with one of my best friends. I had never really paid attention to the yachts berthed along the marina until I saw a massive one. So massive it covered the buildings behind it¹. Its name was Dilbar, and it was the largest private yacht by gross tonnage at that time.

A few subsequent walks only deepened my curiosity. One night, I opened Wikipedia and started reading about the yacht charter business. In the Spanish article, there used to be a sentence that went something like this:

“Latest trends in the yacht charter industry involve implementing chatbots, machine learning, and artificial intelligence to customize the customer experience.”

I looked into applied machine learning in the sector, found nothing, and closed my laptop for the day. An idea was born: I would build an application to help charter agencies recommend the most suitable vessel for a customer based on their previous charter history.

The sensation of building something no one else had built felt amazing, but it came with a great responsibility: you have the power to set a standard, but at the same time, more capable people might be secretly working on the same thing. I had to move fast (I didn’t).

Anyway, I scraped the entire Yacht Harbour yacht database into a SQLite file (if you’re reading this, I’m sorry!) and plugged it into Python, only to decide Java was more suitable for my requirements a few days later.

Fast forward nearly two years, I had a piece of software that would fetch previous vessels from a database using JDBC, and then classify them in two ways:

• For quantitative attributes, such as length, width, number of rooms, and crew, it would look for similar vessels within n standard deviations, increasing n by one if no matches were found.
• For qualitative attributes, which were less relevant but still important (shipyard, interior and exterior designer), it would build a layered graph, traverse it recursively, and create tuples.

The result was a set of numerical attributes with qualitative tuples that were then used to query the database, producing a customized set of new vessels. In the end, the result was presented as a PDF created with HTML, including similarity percentages for each attribute.

The thing is, I did all of this for fun on my legendary ThinkPad T400, running an old version of Eclipse and with no previous knowledge of Java. While I was building it, I thought maybe it was worth showing around, so I cold-emailed several yacht charter agencies and even had calls with some of them.

In the end, I was offered an internship at Camper & Nicholsons, which I turned down for a better offer. The software was no big deal in hindsight, but it has a special place in my heart because it set the foundations for everything else you’ll see on this website.