2025 has come to an end, and it was even more eventful than 2024. I thought I’d do another “year in review” like I did in 2024, but since it ended up much longer than I expected, I decided to split it into two posts. This post serves as a “shorter” summary, with a longer retrospective to follow.
Highlights of 2025
- The Retrieve, Merge, Predict paper was finally published on TMLR in April. What a momentous occasion.
- skrub’s monthly downloads hit about 120k by the end of the year.
- I presented the library at both PyData Paris (video) and EuroScipy in Cracow (video).
- I hosted a bunch of skrub sprints.
- I prepared the material for teaching skrub as part of the Inria Academy program, so that in 2026 I should be able to teach it.
- I obtained a certification as a “Scikit-learn professional practitioner.”
2024 to 2025
First off, let’s revisit what I talked about last year.
Seriously though, every time I try to write something here I spend more time debugging Hugo than I do writing.
As it turns out, I ditched Hugo completely and switched to using Quarto to render the website instead. This turned out to be an excellent decision, as I can now use quarto publish to ensure the page renders correctly. Much saner syntax as well.
Overall, most of the year was devoted to “the paper,” that one article I have been working on for the past two years.
Part of the year was still devoted to the paper, but this time it was accepted, so that chapter is finally closed.
I should start looking for a new position soon enough, but first I’d like to get the paper done with.
In the end, I decided to stay and keep working on skrub until at least 2027. I found the experience to be far more rewarding than my time in academia. It has also been extremely instructive, as I learned so much about programming, hosting events, tooling up, and other skills that I am sure will be useful in the future. I also really like the people I work with, which is another huge plus.
I would say that the year has been very stressful, with burnout waiting for me after pretty much every paper deadline, and some extremely disappointing experiences in academia.
The year was far less stressful, although I did experience some burnout towards the end.
If I have to come up with a theme for 2025, I think it will be something along the lines of “read more sources.”
Nothing of the sort happened. Almost all the sources I stored went forgotten, and my switch from academia to development was almost entirely complete the moment I stopped reading papers.
It’s interesting how I forgot almost everything that I wrote a year ago. I wonder what will happen next year, then.
2025 to 2026
My theme for next year: writing and language. This means dedicating more effort to learning French and maintaining a daily diary. It was a useful habit I picked up but then gave up on completely halfway through the year. I hope to be more consistent in 2026.
I’ll also add “reading more books” to the list, which is something I used to do but haven’t done enough of lately.
Concluding: 2025 in short
Overall, I think 2025 went far better than 2024. My current position feels more rewarding, and I am very grateful to my superiors for shielding me from most of the office politics that occur above my level, though that may change in the future. We will see.
I learned so much through the year, from programming to using tools that make writing programs easier and faster. I got to travel to Poland for EuroScipy, which was a fun adventure, and presented skrub at PyData Paris.
I went from having three colleagues working on skrub, to being alone, to maybe having three colleagues again. What a rollercoaster.
I started a French course to improve my current knowledge of the language, though my relationship with it is still contentious.
Lots of changes, lots of lessons, some nice achievements—overall, a positive balance for 2025. Here’s to an even better 2026!
Happy New Year, and see you in twelve months for the 2026 retrospective. Or, in the next post, where I’ll go through 2025 month by month.