TK
Home

2025 in Review: ML, Books, Career, Writing

• 8 min read
fireworks in the night sky

I'm a bit late, but I still consider it that time of the year we reflect on life, be grateful for all the things we have accomplished in 2025, and plan for the next year quarter.

I like the idea of squashing goals within small timelines to create urgency, fixing procrastination and improving focus. Setting goals for the entire year is too long, and we tend to procrastinate a whole lot. Parkinson’s law is real.

But before moving to the next plans, let's just do a quick recap on the things that happened in 2025.

These are the topics I'll be talking about:

  • Books
  • Career
  • ML Studies
  • Writing
  • Health, Finances & Other Stuff

It's quite similar to the previous review I did in 2023.

Books

I've read 17 books in 2025. Not that much, as I expected, but the quality was way better than the previous years. This is something I reflected on in 2023, when I read 60 books, but most of them were very repetitive, boring, or didn't lead to any really interesting insight I could apply in my life.

This year, I focused on biographies, technical books, and a couple of non-fiction books. This is still my plan for 2026: read fewer, but better books.

I have ML and mathematics books I've been reading the entire year, but I still haven't finished them. I'm making lots of progress on them, though. It's tough, but I'm enjoying this journey of learning through real technical books.

These are the technical ones I've been focusing on:

It's a mixture of experts theory and practice.

The other part of books I'll focus on are: biographies (these are the most fun to read — then having a deep discussion with the Founders Podcast), and non-fiction books in the same vein as How the World Really Works to improve my perspective of the world.

Career

I'm still working on performance engineering, from frontend to backend to infrastructure. In 2024, I focused solely on frontend and learned a lot. In 2025, most of the low-hanging fruits in the frontend were done, so I could focus on servers and infrastructure, with the goal of optimizing our search and listing pages.

Besides that, I've been attending my company's internal AI bootcamp. I'm still in the second month, but I could finish some courses and small projects to apply what I've learned (Python, Pandas, Logistic Regression, Neural Networks, Gen AI/Agentic AI).

These next two months, we need to finish 3 big projects: a regression problem, a classification problem, and an AI Agent model. This will be fun!

One of my goals for the quarter is to be ready to change the title, from Software Engineer to ML Engineer/Data Scientist (if the company has open roles and I'm a good fit for it). Let's see how this is going to play out.

I want to try it out and see if it is a good fit for me, if I like the process of working on data problems, and hopefully, I can learn a lot throughout the whole process.

ML Studies

This is the section I'm most proud of. ML is a really hard subject, and I remember I was very intimidated by the math and the deep technical stuff 2 years ago, but ML is getting more and more ‘comfortable’ for me.

If I look back to the start of 2025, I can't believe how much progress I have made on my ML studies and how much I know now. The scary and exciting thing is: Now I know more about the stuff I don't know. The paradox of knowledge is a real thing.

Here is a list of stuff I have implemented and studied in 2025:

  • A Deep Neural Network from Scratch with Mathematics and Python
  • A Linear Regression from Scratch
  • A Logistic Regression from Scratch
  • A Recurrent Neural Network from Scratch
  • A biology model to predict protein function based on sequences
  • A Computer Vision Model for Chest X-Ray Medical Diagnosis
  • Writing lots of PyTorch code: From Linear Regression to Computer Vision
  • Transformers from Scratch
  • Playing out with Agentic AI and Diffusion Models

All the projects were placed in my ML Research repo. I have a ton of notes in my notebook and in this repo.

Well, that was a year of a lot of work & a lot of fun! I couldn't be prouder. Besides trying to genuinely be a good person, one trait I like about myself is being a hard-working person.

For 2026, I planned to keep working on my studies, but with some other things in mind:

Work on more advanced ML

  • Transformers, Self-Attention
  • Multimodal ML
  • Diffusion Models
  • Deep RL

Papers

  • Read, unpack & review source code from papers
  • Fill the ML knowledge gaps
  • Fill the math knowledge gaps

Research & Applied ML

  • The biggest idea is agentic AI & foundation models for experimentation and simulation in biology
  • I want to reverse engineer each part of this big research idea and work each step at a time: from studies, papers, to the implementation
  • I also plan to research healthcare and medicine problems that could lead to any interesting insights

My studies follow this learning path, mixed with the other one I built.

Writing

The first blog post I wanted to share here is one of the biggest challenges we had at QuintoAndar regarding a performance degradation, the biggest incident we had. It took us a couple of months to work on different hypotheses and experiments to figure out the problem and fix it.

Optimizing Next.js Performance: LCP, Render Delay & Hydration

Because this was the year of studying ML, I built a lot of interesting projects and could write about each one of them.

Here is the list of them:

I really enjoy writing deep technical blog posts. Even though I have written a lot of articles in 2025, I ended the year feeling I could have written more posts.

I was reading 's amazing post on RoseTTAFold, and it got me inspired to work on more deep technical blog posts. There are interesting papers I want to read this year, and I plan to unpack the inner workings of each model.

Hopefully, I can get more work done this year.

Health, Finance & Other Stuff

2025 was the year I invested more money than in any other year in my life. Knowledge compounds, and the financial knowledge I have learned throughout the previous years is helping me do better experiments and improve my portfolio.

Money and finance are not things I like to put too much time into, but I do because it's needed. I spend little time on them per day, but it compounds, and it pays off, literally and metaphorically.

I moved to a new house, and I was getting used to my new neighborhood. In 2026, I feel I'm better adjusted to my new reality, and I can focus on other things now.

Regarding my health, there are a couple of things to say about it:

  • Running: not the best year, but I could consistently run 5k 3 or 4 times a week, which is good, but compared to other years, I feel I degraded.
  • Food: also not the best. Food quality degraded, but it was still healthy.
  • I bought a standing desk, which made my quality of life much better (together with back workouts). I don't have back pain anymore, and I tend to think better when walking around the office.
  • Focus is one of the things I noticed that got worse. My assumption it's due to social media and doomscrolling. Even with this problem, I could get a lot of things done, which makes me think about how many more things I can get done if I fix this issue.

Because my focus is still on working on my ML studies, getting better at it, and changing career, I have a little time left in the day.

I experimented with some ideas that made me realize how much I can do with little time. Every day, I practice guitar for 10 minutes. That's it! Just 10 minutes. I couldn't believe how I could get better with little time, and how the compound effect is powerful for skill building.

I want to try the same experiment to improve my Japanese and work on photography and film.

Regarding finances, I still plan to learn a bit every day and let it compound over time, so not much to adjust here.

Regarding health, I plan to be more disciplined with my food, my running exercises, and to avoid doomscrolling to fix my focus issue.

The plan is also to travel at the end of the year. I will not disclose it now, but I will share it later the places we are going to visit.

Looking forward to another great year of improvement and progress! ✨

Hey! You may like this newsletter if you're enjoying this blog. ❤
✖

Twitter · Github