About me

I'm a Master's student in Machine Learning at Carnegie Mellon University (graduating Dec 2026), working on making large models cheaper to train and serve.

This past summer I was a Research Intern at Microsoft, working on vector quantization for frontier-scale Mixture-of-Experts LLMs — compressing them to under 2 bits per weight while keeping the accuracy loss around 2%. Before that, I was an Applied Research Intern at Tencent, where I built an LLM-agent pipeline for 3D scene generation and fine-tuned small models to replace expensive API calls.

I did my undergrad as a dual-degree student: B.S. in Computer Engineering at UIUC (Highest Honors) and B.Eng. in Electronic and Computer Engineering at Zhejiang University (Outstanding Graduate), where I started doing research on LLM compression with Prof. Volodymyr Kindratenko and in Prof. Tong Zhang's lab.

Recent papers

• SpecDrop — parameter-free category-conditioned routing for modular specialization. First author. arXiv:2608.04084 · code
• Adapt-Pruner — adaptive structural pruning for efficient small-LLM training. Co-first author. arXiv:2502.03460
• RL-Pruner — structured CNN pruning with reinforcement learning. First author. arXiv:2411.06463 · code

I'm looking for 2027 new-grad roles in ML research/engineering and quantitative research. Reach me at bryanw2@cs.cmu.edu.