hello, i'm

Nikita Karagodin

PhD researcher at MIT

precision
recall

I think about why attention works — what makes transformers learn, and what that tells us about learning itself. I work with Yury Polyanskiy and Philippe Rigollet, and I grew up on math olympiads, so watching AI learn to reason feels almost personal.

currently thinking about

tokens as interacting particles on the sphere —
and agents that automate the research itself

Recent

Continuous First, Discrete Later: VQ-VAEs Without Dimensional Collapse May 7
paper
Joining Citadel Securities as a Quantitative Researcher intern Mar 15
life update
Normalization in Attention Dynamics Oct 2025
paper
Clustering in Causal Attention Masking Nov 2024
paper
See all meditations →

Санкт-Петербург

A city of white nights, bridges rising over the Neva, and the quiet beauty that shaped a part of who I am.