Mathematics & Computer Science · UCLA
I’m a Mathematics and Computer Science student at UCLA interested in machine learning, systems, probability, and mathematical research.
I tested whether Qwen3-1.7B’s clean-prompt activations predict answer changes beyond output confidence across held-out ARC and MMLU. Activation models did not beat logit baselines overall, while controlled late-layer patches restored 100% of hint-induced and 92.5% of irrelevant-context answer flips.
I built a scoped ColPali retrieval reproduction and a fused Triton MaxSim kernel that achieved a 2.27× geometric-mean speedup over PyTorch across 12/12 workloads on an NVIDIA L4.
I experimented with RL policies for BTC/GLD allocation, lifting out-of-sample Sharpe by ~14% vs BTC and ~19% vs GLD.
I tested mean-reversion strategies across cryptocurrency futures markets.
I built backend tools for Git-style document version control.
I built a C++ pipeline for ingesting and monitoring telemetry.
I led a team analyzing player-tracking data to measure wide receiver blocking.
Coursework: Stochastic Processes, Probability, Graduate Algebra, Algorithms & Complexity, Systems & Architecture, and Machine Learning.
Activities & distinctions