I'm currently (and will forever be) a student at UMich studying CS, with particular interest in ML, educational technology, language, and blockchain/smart contracts.

I love playing around with ideas — one of my recent projects has been combining arithmetic coding with the output LLM softmax layer for natural steganographic transmission over mediums like ham radios, where encryption is apparently banned.

I've also been messing with zk-SNARKs, and I made a zero-trust version of Battleship in Solidity, which has been a ton of fun.

My meat and potatoes, however, have been ML. I've been super keen on MI and the stuff that the teams at DeepMind and Anthropic have been putting out — lately I've been poking at attention heads and small-model circuits with TransformerLens. I feel it would be super cool if we could understand the inner workings of these models to the point where we can optimize, consolidate, and segment their thought processes — untangling a mess of wires into clean bundles. We're not particularly close to that, but I'm hoping we get there soon.

On the non-technical, I'm kind of all over the place. I'm a huge sci-fi fan (check out The Diamond Age — it's my default niche sci-fi recommendation), especially in the form of short stories. I speak Japanese and enjoy learning about the culture and the history (the Meiji Period is fascinating). I've been trying to get my words out on paper, and so I've been writing on my Substack. It's mainly been a mash of CS, metaphysics, ethics, and political science, but if that interests you, you might enjoy my ramblings. Lately, as a result of my regret for not playing an instrument during high school, I'm trying to learn the piano, and I'm about as bad as I expected. All in the name of trying!