My name is Arul Kolla. I'm a junior at MIT studying Artifical Intelligence and Decision Making in the EECS department. Currently, I'm also a student researcher at the Energy-Efficient Circuits and Systems Group under Kyungmi Lee and Peiyan Dong.

I'm interested in how humans can generalize well from small amounts data, and why models still need large amounts data to not generalize as well. More formally, my current interests include:

# Research

Efficient Generative AI Inference via Nonlinear Kernel Approximation

Arul Kolla, Kyungmi Lee, Peiyan Dong, Chris Mejia, Anantha Chandrakasan

MIT SuperUROP · Energy-Efficient Circuits and Systems Group

Ongoing project

REBOUND: Radioactive Elasticity-Based Objectives for Unwatermarking Neural Decoders

Arul Kolla

6.7960 (Deep Learning) research project, December 2025

PDF · Project

In-Context Learning for Esoteric Programming Languages: Evaluating and Enhancing LLM Reasoning Without Fine-Tuning

Saraswathy Amjith, Michael Wang, Arul Kolla, Jayson Lynch, Neil Thompson

DL4C @ NeurIPS 2025 · FutureTech

PDF · OpenReview

A Sublinear Algorithm for Path Feasibility Among Rectangular Obstacles

Alex Fan, Alicia Li, Arul Kolla, Jason Gonzalez

arXiv preprint, 2025

PDF · arXiv