Hello internet stranger. I am Mrinal Sharma (or as I go by online, MrMarinaru); a second year Computer Science engineering student at Manipal University Jaipur.

My biggest driving factor has been my curiosity. Curiosity about the things that make me wonder, push me to figure out more about the unknown around me. To understand how and why it works. That same curiosity fuels me to constantly try and improve in the fields I love. I like to take things apart and view them inside out. Whether I understand it or not, the mere fact of looking at something unknown that I may not yet understand, makes my brain squirm with fascination; a fascination to look more.

I’ve been programming for around four years now (five, if willing to include those childish Scratch scripts), mostly in Python. When I started, I was simply messing around with the concept of programming, making the most random ideas I could think of, come to life. I then gradually moved into working with data — learning MySQL, pandas, Matplotlib, NumPy, before diving into Machine Learning and Deep Learning. These days I like to experiment with libraries like scikit-learn and PyTorch to build small but meaningful projects that help me understand the math behind what’s happening under the model's bonnet.

To understand what I am building, I slowly have noticed a shift inside me going from using LLMs to hallucinate my code to trying my level best to make the things I want to use, myself. It provides me a layer of understanding I could have never got if I hadn't tried the task myself. Programming's joy for me lies in the problems I solve, not just the results I accomplish. The mere fact that I developed a typing speed of 80 WPM over large blocks of text, goes to show how much I love to type out my own code.

So far, I’ve worked on a bunch of ML projects — like a plant image classifier using transfer learning with EfficientNetB3 & ResNet50, a yield prediction model for a real-world agricultural dataset, and even a race lap time predictor for MotoGP data. I've employed various algorithms for said works, including but not limited to Random Forest Regression, Decision Trees, XGBoost, ADABoost & other techniques. I’ve also played around with RBMs and Autoencoders built from scratch in Theano; later porting them to PyTorch just to understand how deep networks really work.

For me, Understanding my concepts takes a heavy priority. Therefore in order to achieve that, I heavily lean upon books written by accomplished mathematicians and researchers, to clarify my concepts at a significant level. For example, I can't fully express the wilds I felt as I was reading about RBMs and found that they employ minimizing an energy fn to calculate the probabilities under the hood. Trying to wrap my head around Monte Carlo Markov Chains, Contrastive Divergence, Back Propagating and other concepts I hadn't encountered before felt like a new world opening for me. And when I realised this wasn't even the state of the art techniques but just relics of the past, atop a feeling of disappointment; I found in myself the feeling of having found a world so large, I could drown in the sheer amount of marvels waiting to be explored.

Outside of code, I’m just a regular guy; I game, I watch anime, I try to stay fit and live my life without leaving regrets for future me. This blog is where I’ll be sharing what I learn, break down my exploits, and maybe write about random thoughts that connect tech, creativity, and life as an engineering student.

If you’re into building stuff, learning conceptually sound topics, or just reading about the mess that is trying to become a better developer, you’ll probably feel right at home here. Maybe even reach out to me while you are at it!