From Naval Architecture to AI: Why I Changed Everything
This is not a technical post. It is an account of how I ended up studying Data Science and AI at the University of Liverpool after four years of Naval Architecture and Ocean Engineering in Korea.
TL;DR: Studied Naval Architecture in Korea, had one Python class that changed things, enrolled in an AI bootcamp, got accepted to MSc Data Science & AI at the University of Liverpool. This post is about why.
1. Naval Architecture
I studied Naval Architecture and Ocean Engineering at Chonnam National University. The degree covered fluid mechanics, structural mechanics, ship hydrodynamics, and thermodynamics - a rigorous engineering curriculum by any standard.
I did not dislike the subject. But I never felt drawn to it either. It was a degree I had chosen, studied, and completed - without ever feeling the pull to go further.
The moment that changed things was unremarkable on its surface. One course involved Python programming - nothing sophisticated, just basic scripting and numerical methods applied to engineering problems. But somewhere in the middle of writing that code, I had a clear and simple thought: I want to do this for a living.
Not naval architecture. Not ship design. This - writing code, building things with software, making something work from nothing.
2. The Decision After Graduation
After graduating, I had two options. Find a job in naval architecture and ocean engineering, which was the expected path, or do something about that feeling I had in the Python class.
I chose the second.
I enrolled in a government-sponsored AI and machine learning bootcamp in Korea. The programme was intensive and vocational - designed to get people employed in tech as quickly as possible. The curriculum moved fast and covered the basics of deep learning.
The more interesting learning happened outside the formal curriculum. The computers provided for the programme had limited resources, and fine-tuning a model on constrained hardware is not something the course covered directly. I started looking into memory-efficient training methods on my own - how to reduce the memory footprint of a model without sacrificing too much performance, how to get something working on hardware that was not really built for it. That problem-solving process led me to LoRA and quantisation techniques, and eventually to deploying the results as an API. None of that was assigned. It was just the only way to make the project work.
And I found it genuinely enjoyable. Not in the way that some things are enjoyable when they are going well, but in the way that makes you stay up later than you planned. Reading documentation, debugging something that should work but does not, eventually getting it to work - I did not have to force any of it.
The specific things I found most engaging were reading papers and implementing them, and taking a trained model and deploying it as a working service. The gap between a research idea and a running system is where I wanted to operate.
3. The Limit of the Bootcamp
The bootcamp was good at what it was designed to do. But it was built around employment outcomes, not depth of understanding.
I started noticing the gap. I could implement things, but I did not always understand why they worked. I could fine-tune a model, but the theoretical foundation was thin. The papers I wanted to read were in English - and reading ML research in a second language at speed was difficult.
Two things became clear. I needed to go deeper into the material. And I needed to do it in English, because that is the language the field operates in.
Both pointed in the same direction.
4. Why the UK
I looked at graduate programmes in several countries. The UK made sense for a straightforward reason: a one-year MSc meant I could go deep quickly, and English immersion would happen by necessity rather than by effort.
I applied to the University of Liverpool’s MSc in Data Science and AI. The programme covered the areas I wanted - machine learning, applied AI, computational intelligence, databases, mathematics and statistics. It was also taught entirely in English, which was the point.
I arrived in Liverpool in 2025.
5. What Changed
The difference between the bootcamp and graduate study is not primarily the content, though the depth is significantly greater. It is the mode of engagement.
In a bootcamp, you implement what you are shown. In a research-oriented programme, you read the original papers, understand why decisions were made, and build from that understanding. The RAG system, the RL racing agent, the fine-tuning experiments documented elsewhere on this blog - all of them started from reading primary sources and working through the details.
That is what I was looking for in the Python class in my third year of naval architecture. It took a while to find the right environment for it.
6. Where I Am Now
I am currently in the middle of the MSc, due to complete in September 2026, and looking for ML engineering or data science roles in the UK. The projects on this blog represent what I have built during this period - not coursework submissions, but things I built because I wanted to understand them properly.
Naval architecture gave me an engineering foundation and a tolerance for technically dense material. The bootcamp gave me practical instincts. The MSc is giving me the depth I was missing.
The transition was not dramatic. One Python class, one decision, a few years of working toward it. That is roughly how it went.