Kushal Thaman

I’m Kushal (pronounced /kʊʃəl/), an undergrad at Stanford studying Computer Science. Previously, I worked on the science of neural network training at STAIR, adversarial attacks at CAIS, and now think about in-context learning & build new LM architectures at Stanford NLP, supervised by Prof. Chris Manning and Prof. Chris Potts.

In my spare time, I like to do photography to make the world feel a bit more mythic, swim, think about culture & endogenous growth theory, and counter my perpetual onism by hiking mountains.

Reach out if you’d like to chat about anything you think we’d both enjoy!

news

Jul 17, 2024 I have been awarded an Emergent Ventures grant by Tyler Cowen.
May 11, 2024 I presented my co-first author interpretability paper Incidental Polysemanticity as a poster at the Representational Alignment workshop and the Bridging the Gap Between Practice and Theory in Deep Learning (BGPT) workshop at ICLR 2024 in Vienna, Austria.
Feb 25, 2024 I presented our paper on Probabilistic Syllable Structure at the 42nd West Coast Conference on Formal Linguistics (WCCFL) at Berkeley, CA.
Feb 3, 2024 I am a Teaching Assistant for the course CS21SI: AI for Social Good in Spring 2024. Come hear me talk about RL and unsupervised learning!
Dec 15, 2023 First day at the Center for AI Safety, where I’m working on vision adversarial attacks with Dan Hendrycks.
May 21, 2022 I have been awarded the Atlas Fellowship.
Jul 1, 2021 I attended the national orientation & selection olympiad camp for the International Olympiad in Astronomy and Astrophysics (IOAA), organized by HBCSE.
May 28, 2021 I was part of the five person team that represented India at the International Young Physicists’ Tournament (IYPT).
Dec 17, 2020 I won the Honorable Mention award at the International Physics Olympiad (which was held as a distributed event due to COVID-19).
Jun 16, 2019 I achieved national rank 5 in Physics at the STEMS exam and subsequently attended their camp on Physics, Mathematics and Theoretical Computer Science at the Chennai Mathematical Institute.