Bruno Eilliar's Blog

New year, new me... is what they say. As with every new year, I take some time to set new goals, both professional and personal, for the year that comes. In 2025 I setup to quite a few things. But one thing on my study plan are proving to be the one that I'm most engaged with: learning game development and C++.

You see, C++ was something that I learned at university, but since I started to work at CargoX, never more I touched a single line of code in C++, relying entirely on Python to do my job as Data Scientist. But everything started to change last year, around November, when in order to have a reasonable excution time for a hyperparameter search for UMAP and HDBSCAN I had to install and use RAPIDS. Then I decided to understand what RAPIDS was using under the hood to be able to run algorithms in GPU, that was when I started to understand the way CUDA works. Specially through this amazing tutorial. I ended 2024 doing some simple CUDA experiments then I noticed that a lot of C++ was strange to me. Python made me very lazy in the past 7 years. Then I decided: I need to re-learn C++. I started to take a look on PyTorch C++ API to try to replicate basic neural nets on C++. But then I started to get frustrated, specially because I knew how to code the same nets in very few minutes using Python API instead of C++, the progress was really slow and boring. Then it occurred to me: since C++ is widely used for game development, why not learn C++ while learning game development? (as a gamer this made complete sense to me).

So it's 2025 and I bought Beginning C++ Game Programming - Third Edition and started to code the games. It ain't much but it's honest work:

It proved to be much more fun to learn something new when you build something that is also new to you. Coding for data science tasks is really different then coding for games. Now I have to understand game loops, work with user inputs, events, etc. It's a lot of best practices that I hope find some usefullness on my job as a Data Scientist.

The book uses SFML as framework for game development, making it very easy to put images on a screen and make them move, play some sounds and window management. But then I started to wonder, "how the fuck is this transforming data stored in computer's memory into images on my screen?". Then I started to research... and research (one can say this is a trait of any Data Scientist, wanting to understand how an algorithm works). Then I came accross something that is changing my mind: lectures on computer graphics by scratchapixel and it's beautiful 🤩. So now I'm stuck, 1h to 2h a day working my way through the lectures and understanding that the transformation matrices that I used in robotics are basically the same stuff used to project stuff from a 3D simulated word into a canvas (ok, it's an oversimplification, but the concepts are the same). That is why math is so beautiful, different applications, same divine language.

So that is my goal for 2025: become better at C++ through computer graphics and game programming. Maybe even launch a demo game on Steam 😂... ok, let's be real, maybe on itch.io (I've the feeling that it's easier there). I even built a plan:

And to keep track of my progress I decided to revamp my GitHub page and transform it into a blog. So here I'm, 2025, after a 2024 that was a roller coaster I decided to re-learn a very old language (sorry Rust, you will have to wait) building stuff that are completely not related to my job. I hope this proves to be something worth the time, having fun along the way is something I would be satisfied with. To finish this post, here is some 3d projections that they teach on scratchapixel: