About
I am a Member of Technical Staff at Perplexity, where I work on computer-use agent RL training.
I earned my M.S. and bachelor's degrees at the University of Illinois Urbana-Champaign (UIUC), where I worked closely with Prof. Minjia Zhang on PyTorch compiler optimization.
I also contribute to open source, including SGLang, Miles, TransformerEngine and PyTorch.
Research interest: efficient computer-use agent training.
News
- Aug 2026I joined Perplexity as a Member of Technical Staff, working on post-training.
- Apr 2026Our blog Introducing AutoSP has been featured on the PyTorch Official Blog.
- Jan 2026Our paper AutoSP has been accepted at ICLR 2026.
- Jan 2026I joined ByteDance Seed as a Research Intern.
Experience
Perplexity
- Develop fully-async RL with SGLang rollout; shipped the training observability dashboard.
- Build the low-precision SFT & RL training pipeline with NVFP4 and MXFP8.
- Develop an end-to-end debug workflow for large-scale SFT on GLM-series models.
ByteDance Seed
- PyTorch compiler optimization.
Selected Publication
ICLR 2026 · *Equal contribution PDF
Blog
PyTorch Blog · Apr 2026