2025 Wrapped

git commit -m "Year of pure Machine Learning"

Welcome to my first-ever year-in-review. To be honest, I'm much more comfortable writing code than writing about myself, so this is all pretty new to me. I've never done a public recap before, but after a big year of diving into Python and Machine Learning and where I'm headed next.

Images of the Year

10th Aug 2025: Solid day at GDG Ahmedabad learning all things Google Cloud. Good vibes, great friends, and a lot of new cloud knowledge.

16th Dec 2025: Presentation day for SSIP 2025 (GLS Law Laboratory (Container)).

Skills and Deep Dives

This year, I intentionally went back to the basics, especially pure Object-Oriented Programming. I moved away from writing mostly linear, procedural code and started thinking in terms of systems and entities. Treating everything as an object helped me reason better about state, behavior, and how different parts of a system interact, which is crucial when building things that need to scale.

A large part of my learning came from structured academic material. I spent time with IIT Madras YouTube Playlist and the Cornell University Spring course. These were not quick tutorials but deep academic dives that pushed me to understand why things work, not just how to use them.

Alongside this, I revisited Introduction to Statistical Learning with Applications in Python as a core reference, especially for topics like resampling methods and regularization. Combining Cornell's theoretical rigor with ISLP's practical framing helped me move beyond just calling library functions to actually understanding loss optimization and the bias-variance tradeoff. Also earned certificate on NPTEL in Introduction to Machine Learning.

By the end of the year, I felt a clear shift in how I think. Instead of asking which model or tool to use, I started asking what assumptions I am making and how those choices affect generalization.

Learned by Solo

A lot of this learning happened alone. No fixed deadlines, just me, the material, and the discomfort of not understanding things immediately. Studying this way forced me to sit with confusion longer, revisit the same concepts multiple times, and build my own mental models instead of copying solutions. It was slower, but it made the understanding stick.

I also used AI to make sense of a few tricky topics. Turns out having something that never gets tired of dumb questions is incredibly useful. Totally worth it! 😁

Projects

This year, I also co-founded Colrnx with Vashisht Brahmbhatt and Daksh Patel. Most of the work so far has gone into shaping the idea, defining the direction, and building a solid foundation rather than rushing to ship. It's still early, but the plan is to launch it sometime in 2026, and I'm excited to see how it evolves from here.

Things that didn't work

I tried getting into the creative side this year, things like content writing and video editing, but it didn't really work out. Nothing much to elaborate it.

Carrying this forward

Going ahead, I want to go deeper into machine learning instead of spreading myself thin. This means spending more time with tools like scikit-learn and PyTorch, getting better at data preprocessing, and understanding how real-world data is shaped before it ever reaches a model. I also want to slowly build the habit of reading research papers, not to rush results, but to get comfortable with how ideas are explored and validated.

Benchmarks/Stats

That's it! I'm glad I took the time to write it down.