About

I’m a second-year Computer Science student at the University of Waterloo. Previously, I worked full-time on AI system evaluation and developer infra at one stealth startup, and part-time on data processing at another.

Deep learning in NLP is my main area of interest. I am looking into the extent to which LLM token embeddings and hidden states are capable of expressing outputs and meanings. I perform small explorations on these topics, which you can find on my Github.

On the side, I also play jazz on the alto saxophone and flute.

You shall know me by the company I keep. Barry Yu - more soon…

News

• [Sept. 2026] I’m super chuffed to be joining UWaterloo’s CompLING Lab part-time to work on my passion project!
• [May 2026] I’ve started a temporary position as Software Developer at a stealth startup!

Blog

1. 2026-08-19
   Notes for proving the Softmax bottleneck

   Why language models can't perfectly model language

2. 2026-07-02
   Binet's Formula w/ Power Series

   Deriving a closed-form expression for the n-th Fibonacci number

Powered by Jekyll; based on the Minimal Light theme.