A report on my 2024 self-reflection and checkpoint, and how I want to move forward with my career growth in 2025
Introduction
If you know me, you probably randomly wonder why I stopped posting on LinkedIn when you come across my profile. I had written a self-reflection for the first half of 2024 back in July 2024 over there as part of my self-accountability (I have a post on that here on this site as well), and I had some things I wanted to achieve by the end of 2024. However, while I have been meaning to write a follow-up, I never got around to it, although that doesn’t mean I haven’t been working on things :) in fact, I’ve never been more busy in my life than I was in the second half of 2024: learning new things, creating a timeline for my career growth, and so on. This meant my priority towards updating my blog or writing on LinkedIn became more of a lower priority.
Of course, I did mean to do a year-end self-reflection post for 2024 back in December 2024, but one thing led to another and I ended up not writing one (until today, at least). As of writing this, it is March 16, 2025, and although I believe this is a bit late, I still want to share my thoughts on what I did in 2024, what my 2025 resolution is, and what I’ve been up to in 2025 so far.
2024: A Year of Growth
So, if you read my post on the first half of 2024, you’ll know that I had some things I wanted to achieve by the end of 2024. So, I’ll firstly go through each of them and see how I did, and I’ll also share what other things I was able to do in 2024.
Looking Back at My 2024 Objectives
1. Reach an intermediate level of mastery in JavaScript, TypeScript, and React.js.
Not an exact method to quantify this, but I do believe I’ve gotten quite okay with React.js, and I’m way more familiar with JavaScript & TypeScript than I was before. I use React.js at work pretty much every day, and I’ve learned a lot about the best practices while building a site using it.
2. Travel around Kyushu! I want to return to Beppu to try the green curry at APU again.
I went to Fukuoka and Beppu! I could not get enough time to experience the green curry at APU, but I had a great time visiting restaurants and cafes I used to go to as a student.
3. Complete the challenging Neural Network- Zero to Hero course series.
I completed 3 courses of the series (micrograd, makemore’s first video, and the ‘let’s build a GPT’ video). I paused the series after that since I started working on other things, but I plan to pick it up again soon.
4. A difficult one, this, but I also want to build at least 1 large language model (LLM) of my own from scratch by implementing Sebastian Raschka's 'Build a Large Language Model' book.
I actually completed this one! I managed to build my own GPT architecture by using this book (and Karpathy’s course) as a reference, and used it with pre-trained weights from OpenAI’s GPT-2. Also tried training it on small dataset from Hugging Face, and that was a bit of a disaster, but it was a lot of fun! I have written about it here on my projects section as well.
5. Make more friends!
I’ve made a lot of friends in 2024, and I’m still making more!
A Year of Learning and Shift in Mentality
2024 was pretty much a ‘get your head down and grind’ kind of year for me.
If I have to look back at 2024 as a whole, I would say that this has been the most important year of my life so far in terms of the shift in mentality from being someone who liked the idea of working hard and grinding, to someone who is now actually working towards their goals every single day.
I’ve obviously known that I have the willingness to learn difficult things and that it takes a lot of effort, discipline, and tenacity to learn them, but one thing that has been a curse for me is the mentality of wanting to have a perfect timeline and clear path. I believe 2024 was the year where I finally got rid of that mentality and focused solely on just DOING, meaning I no longer waited for the perfect thing or a perfect roadmap. Well, this did come to me pretty late even in 2024, but now that I’ve finally gotten there, I feel like I’m finally able to move forward with my life and career. Wanting perfect timelines and clear paths leads to overthinking and overplanning, and I’ve realized that I can’t do that anymore. Doesn’t matter what it is, doesn’t matter if it’s not perfect, doesn’t matter if it’s not a clear path, and doesn’t matter if it feels too difficult or if it doesn’t feel like I’m building things brick by brick. I need to just DO, and I need to do it now.
In other words, I have become a firm believer in just getting my hands dirty and piling up the hours without caring about the efficiency of the work, since it’s just all piling up the time (and in turn, the effort and the learning) and ensuring that I’m always getting better every single day. It can be a bit of a struggle at times to not care about the efficiency of the work, but I’ve found that it’s a lot more rewarding in the long run. I had listened to Andrej Karpathy talk about something like this some years ago, and I think I resonate with this a lot more now than I did back then (paraphrasing):
‘It doesn’t matter what methods or tools or frameworks you use, what matters is that you’re piling up the hours. Do something, iterate iterate iterate, ensure you record your hours, write about what you did, post it, and keep repeating. The 10,000 hour rule.‘
2025: A Year of Metamorphosis
I’ve put a lot of foundations in place in 2024, and as I continue to learn more and more, I believe it’s getting to the point where I should start sprouting my wings and start flying.
Although not exactly a strict set of resolutions, I do have some goals that I want to achieve in 2025. However, unlike how pragmatic I was in 2024, my goals for this year could be regarded as a bit more of a YOLO-ish ‘I want to be there’ with no strict timeline or roadmap. This could get dangerous — especially if one tends to get distracted or cannot hold themselves accountable unless they have a pragmatic sense of clarity in their timeline — but I’ve never been more accountable and focused than I am right now, and combined with my strong sense of belief in my mindset, I believe I’ll be fine.
Looking Ahead to 2025
1. Proficiency in Neural Network Architectures
I want to become proficient in the architectures and the overall ins and outs of all the state-of-the-art models and techniques in the field of LLMs. I’m currently already underway with the Transformer architecture (blog coming up soon), and I’ll soon also be diving deep into Group Relative Policy Optimization (GRPO) and try to replicate a lot of architectures behind some notable models (such as the ones by Mistral AI, LLaMA, DeepSeek, and more).
2. More technical blogs
I am targeting to write at least 1 technical blog post per month, and I’ll be sharing my learnings and experiences on this site as well as LinkedIn.
3. More AI projects
I’ll figure out the specifics whenever I have to, but I want to build as many AI projects as I can. Doesn’t matter what kind of projects they are, I just want pile up the numbers and ensure I’m always getting better every single time. My current focus is, to the extent possible, building as many neural network (/LLM) architectures, optimization techniques, etc.) and reinforcement learning projects as I can.
4. Replicate as many papers as I can
This could intersect with the above goals a little, but I also want to replicate as many papers as I can. Currently, I’m thinking of replicating 1 paper per month, and I’ve already pretty much completed my first paper: Attention is All You Need (reflection blog coming soon).
5. To not forget my friends
I’ve already started doing this, but looking back at how much I pretty much stopped talking to my friends from university in 2024, it felt a bit bittersweet. I did make new friends in 2024, but I was pretty much a caveman when it came to catching up with my friends from university and so on. Therefore, I decided (back in January 2025) to start catching up with as many of my friends as I can and stop getting caught up with life. I’ve already started catching up with some of my friends, and I’ll continue to do so in 2025. If you’re reading this, and you wanna catch up, please reach out! If we’ve been good friends back in the day, I’ll most likely reach out to you soon anyway, but if you are reading this and want to catch up, I’d be really happy to receive your message. You know what, if you’ve actually read this far, I’d be even happier to receive your message because I truly do appreciate you going through this long ramble of mine.
6. Overall, become really good at Machine Learning (ML)
Well, this is a pretty generic sentence, but I want to get the feeling that I can pick up any new ML technique or tool and feel confident about using it. For instance, I’ve been using PyTorch pretty much every day, so I’m pretty confident that I will be able to get much better at it. I also want to pick up CUDA and Triton some point later this year, and I’ll see how it goes! This also includes getting more familiar with ML Systems and Infrastructure and so on. I do need to remind myself that I should not dwell too much on just the tools by trying to obsess over any individual thing too much and instead prioritize familiarizing myself with those said tools or frameworks (or even languages) whenever I have to learn them (especially during the process of building my own projects). In other words, quick iterations and quick learning; it’s not about learning the tools; tools are just a means to carry out an objective, and it’s the latter that I should focus more on.
7. Build build build, and hope that something sticks and becomes a side hustle
I have tons of ideas for apps. In this golden age of tech with the ever-advancing tools and technologies in the field of AI, anyone can build whatever they want effectively as long as they want to. I just so happen to be someone who is up-to-date with the latest tools and technologies, so I’ll definitely be building a lot of things. There will be lots of failures, but I’ve seen people generate $100k/month with just a simple plane flying game; it’s such a fun time to be alive! Even if it doesn’t work out, I’ll still have a lot of fun building it (and learn a lot in the process)! It’s the year to build, fail, and continue iterating.
8. Get a Machine Learning job
I’ve been working harder than I have ever worked before, and this has all been leading up to this. By the end of 2025, my quintessential goal is to get a role in Machine Learning. I am targeting either a ML Research or a ML Engineering role, and I have started reaching out to lots of professionals in Japan stating my goals and intentions. What I lack in experience, I believe I make up for in enthusiasm, willingness to learn, and a strong belief in my mindset that I will always learn and improve. Just need that one shot to get me in the door.