I like problems where there's a number at the end that tells you whether you were right.
I build small, measurable systems, detection models, forecasting pipelines, tools that fix a specific annoyance, and I like it best when the result comes with a number attached.
About
I'm a B.Tech CSE student specialising in AI and machine learning. Most of what I build comes out of hackathons or from running into a problem I wanted gone.
If something breaks, my first instinct is to build the replacement rather than find one. That's how ShellPDF happened: my file converter stopped working, I needed my professor's slides as PDFs, and three hours of writing a tool won out over five minutes of doing it by hand. It usually pays off by the fourth or fifth time.
These days I split my time between a research paper, whatever hackathon is on, semester exams, and trying to make the things I've already built slightly less embarrassing.