Roberto 2027

16 Jan, 2026

Wait, what? 2027? But 2026 just started! and I’ve recently made goals for the year…I know, but what better way to speak your desires into existence than sharing them with the world.

Inspired by Celeste (I resonate a lot with your writing) and Lydia’s “open sourcing your thoughts.” The goal of [Name] 2027 is to visualize who you want to be by that point, write out your goals and work endlessly towards them, reflecting along the way.

2027 Desires

• Form a clear understanding of my career trajectory

• Advance intelligence (my own + the world’s)

• Consistent expression of ideas and work to the world

How I Will Get There

This section serves as an explanation to all the desires I listed above. As I’m writing this, I’ve realized it will most likely be unorganized but I’ll attempt to be structured.

Career

I started 2025 fully convinced that I wanted to work on full-stack software development. I had set my goals, and at no point between the start of the year and the end of the summer did I think to question them. Even as I started my Master’s in August, I was tunnel-visioned into this idea of myself that I believed to be my destined path.

One of the main reasons as to why I decided to go for my Master’s was to go deeper into research, machine learning and deep learning. My senior capstone project was what kickstarted this pivot, because through that work I realized I enjoyed the application of ML way more than traditional software.

So around the start of my Master’s, I started looking for internships within the AI/ML field, confident that I had what it took. Unsurprisingly, I found little to no success during this period, only landing one interview. I prepared for this interview for weeks like my career depended on it (it didn’t). In this interview, I realized that what I thought I knew about ML was actually nowhere close to what I had to know. In hindsight, that failure has derailed me towards a path that I enjoy so much more, and all it took was one interview.

This realization was around late October/early November of 2025. A week or two later, I decided that instead of dwelling on this failure, I would completely re-invent myself. So I honed in on one area of ML that I loved (computer vision) and explored the model that I resonated the most with (Meta’s SAM). I wanted to build it, because I wanted to understand all the ins and outs and prove that I understood it.

The approach I took for this quest was simple: Read the SAM research paper, understand the architecture, break it down into small digestible pieces and work up from there. If I did not understand a certain portion of the paper, I would go to the citation and paper that preceded such section and read it. Given that I had so little understanding of anything, I ended up recursively working my way down to recurrent neural nets, one of the most primitive architectures that there is.

This is where I realized that, in order to go anywhere within this field, I had to build a proper foundation before moving forward. This foundation, rooted in math, deep learning understanding and programming, would be what helped me slowly build up my knowledge up to modern day standards. I still have so much left to learn, and I haven’t built SAM yet (starting very soon), but I’ve seen some serious improvements in the span of just 1.5 months.

For example, during my recursive traversal of papers, I ended up reading 2014 Sutskever et al. as the paper that preceded the Transformer architecture. During my first read back in November, I had no clue what an LSTM was, what an encoder/decoder were, etc. After my from scratch (and by hand) implementations of an RNN, an LSTM and even the paper itself (Github link), I revisited the paper, and had no trouble understanding any concept within it. I felt like everything up to that point had clicked, and all the videos I spent watching, papers I spent reading, and math I spent writing came together.

All this work in the past month or so has further reinforced my desire to work in the field of deep learning/machine learning, but I’m yet to find the specific niche that I want to dive even further into. For now I think it is vision models, but that is subject to change. It has helped me realize that certain topics that seemed ungraspable are actually not, and that its all about putting in the work and dedication. It makes it so much easier that I thoroughly enjoy and dream (literally, these things keep me up at night) about the possibilities of ML.

As I stated previously, I still have so much left to learn, but that’s okay. I love learning, and I will have to learn for life to get to where I want to be. As for how I will form a clear understanding of my career trajectory, I will have to take this course of action:

• I need to figure out whether I go for a PhD or not. This decision is quite monumental given that it weighs in so many things (time, money, location, etc.). So far, I am undecided, and I maybe want to explore the idea of going into industry research (Research Engineer, AI Engineer, or any adjacent roles) to where I can still learn and advance intelligence prior to taking that leap.

Advance Intelligence

When I talk about “advancing intelligence,” I am referring to the concept of gaining new knowledge and forming new manifolds rather than floating around within an existing one. I aim for this knowledge to be absorbed by myself, but also to be shared with the world. Before 2027, there are many ways in which I could do this:

• Master’s thesis

  • This should give me proper insights into whether I really want to do a PhD or not, and will show me if I’m capable of coming up with novel ideas + cool hypotheses. I’m really excited for this.

• Neuroscience research (ongoing)

  • My ongoing research at the intersection of neuroscience and ML could be cool to continue exploring. The concepts within the neuro world that could be applied to ML are uncountable, and I would love to keep going down that rabbit hole. Side note: I recently saw a position which explores this intersection to the nth power and would really love to be part of that…

• Summer/Fall Research

  • It is still only January, so I still have the chance of finding cool research for the summer. I will do my best at reaching out to professors, show my commitment to not just the work but the process, and hope to work on cool stuff. I’ve also explored the idea of fellowships, for which I will most likely have to start writing for soon if I want to explore those.

Expression of Ideas

This section will most likely be significantly shorter than the previous, albeit more important than the previous ones.

I have always been a little reserved on what I share on social media, whether personal or professional. I used to think that this was fine, because at the end of the day my skills or work would speak for themselves. Unfortunately, in 2026 this is no longer the case, and if you don’t put yourself out there, no one will find you. Towards the end of 2025 I started a blog and began being more active on YouTube, X, Reddit, etc. and I found so many cool people and ideas. A lot of the reflections in this post, and even the post itself, are owed to people I’ve interacted with and things I’ve read online.

The more I invested (posting quality content, quality replies, etc.), the more up to date and rewarded I felt. Additionally, I saw myself gaining more traction in certain areas to where I could further leverage such in the future.

My goal for this desire is simple: to keep sharing. If I don’t share, no one will know, and I will not remember. That’s why in order to reach Roberto 2027, I will journal, blog, post on YouTube, LinkedIn, and everywhere I can think of. This encapsulates my true passion: to share the little I know.

Possible Points of Failure

Keeping in mind the parts of a system that could probably cause the most damage is a good way to prevent that said damage from happening, which is why I will expand on what I believe to be my biggest possible blockers to achieving the aforementioned desires.

Noise

Despite working on optimizing algorithms for de-noising Drosophila neuron imaging, I am quite susceptible to noise myself. Especially since I’ve started putting a bigger emphasis on platforms like X, I’ve found it really hard to blur the line between good/bad content. There is so much content out there, whether AI generated or not, that makes you rethink your skills and your persona.

I’ve slowly realized that this will always be a thing, and it was a thing even before social media. You might feel behind because a classmate of yours achieved something great, or even worse because a 14 year old created a $100B company, but at the end of the day you just have to block it all out. In order to reach my desired self, I will have to place a filter over everything that I consume, and tread accordingly.

Priorities

Prioritizing the important work for me is also an anticipated blocker. I find it more troubling to form a linear path from where I currently stand towards a said goal rather than mapping out the goal itself.

For example, in my ongoing deep learning journey, I’m in the process of implementing a Transformer, which I want to properly and deeply understand before moving on to Vision Transformers. I’m implementing it by hand, which is so much harder and time consuming than just using autograd and the built-in PyTorch nn.Modules. This is theoretically slowing me down from my ultimate goal, but I don’t know whether to call it a blocker or not, because at the end of the day the learning goes beyond just surface level.

Funny enough, this post itself could be categorized as a distraction, because I’ve spent way too much time writing this, but my mind probably won’t stop thinking about it until I post it. For balance, I need to not just prioritize work and research, because as much as I wish, I’m not a robot. I should parallelly prioritize health, spending time with the ones I love, and finding a good balance. That is how I’ll stay on track.

Validation

This point is emphasized a lot more on the technical aspect of things, but I feel like I struggle internally with the thoughts of: “what if I’m late”, “what if I won’t know how to”, “what if I’m uncapable,” and so many others that I’m sure are shared sentiments among humans.

Lately, I’ve been approaching these thoughts with the mindset of: we will cross that bridge when we get there.

Realistically, I will never feel on time for research, I will never feel prepared, and will never know everything that I need to. That is okay! I just need to find a way to prevent these thoughts from slowing me down and taking me off the track that I’m currently on, because every day you can be a new and improved version of yourself. That is something that I want to continually compound until Roberto 2027.

Conclusion

If you made it all the way over here, I appreciate you taking time to read this blabber. Writing down my thoughts in order to reflect on them and hold myself accountable is something I love doing, and this is just an amplified version of such. If you feel like this read helped you in certain ways, please don’t be shy and comment below, I’d love to chat.

Footnotes: