What I'm up to lately, a little about me, where I've been, and the occasional thing I write.
Hi, I'm Arnav — a curious undergrad who likes taking models apart to see why they work. Outside research I read widely and tinker constantly.
I'm a B.Tech student in Energy Engineering at IIT Roorkee (2024–2028) and a core member of the student-led Data Science Group, where I research multimodal generative models and vision-language reliability. I've implemented transformers and CNNs from scratch and placed in several Kaggle and ML competitions, with rigorous EDA driving architecture and feature choices.
Across my work, one thread recurs: I care why a method works, not just that it does — whether that means recovering the denoising steps distillation throws away, or showing that an apparent hallucination fix is really a distributional artifact.