I study combinatorics, optimization, and pure math at the University of Waterloo, with a minor in computer science. I’ve interned as a software engineer at Achievers and Intuit, and I build things on the side. I’m seeking new-grad software engineering roles for 2027.
Runs a 27B LLM entirely in 16 GB of VRAM on a consumer AMD GPU and tunes it to drive a coding agent: speculative decoding took decode from 33 to 59 tokens/s, and custom HIP kernels sped up ternary-model prefill by 53%. Every change is backed by checked-in benchmarks.
Interactive rules guide and move checker for Sheng Ji, the Chinese trick-taking card game.
Semantic search over 4,800+ movies: describe a vibe in plain language and get matches ranked by sentence-transformer embeddings, served from an in-memory NumPy vector store.