I'm a third year undergraduate at UC Berkeley. I'm interested
in making LLMs cheaper and more personalized through routing, persistent-memory, and cross-model caching.
I enjoy playing badminton and writing.
May 2025
Routes to the best and cheapest LLM based on each individual query. 50-90% cheaper than best model with 2-3% quality loss.
Link
June 2025
A way to visualize GPT/Claude conversations in interactive 3D space. Built with Dev and Neel.
Demo
