PhD candidate.
I'm a Ph.D. candidate in Mechanical & Aerospace Engineering at Princeton, working with Egemen Kolemen in the Plasma Control Group, joint with PPPL. I work on foundation models for fusion data, the real-time signal pipelines that feed them, and the control systems that turn both into something a tokamak can actually use.
I started the Ph.D. in March 2025 after finishing my M.A. at Princeton MAE. Before that I studied physics and cognitive science at UCLA, founded the American Nuclear Society chapter there, and co-founded the California Neurotech Conference (CNTC).
I'm a 2026 Princeton AI Lab Graduate Student Fellow. I've led a real-time ML control experiment at the DIII-D National Fusion Facility and sped up the ORBIT-GPU plasma simulator by 20× on CUDA. Outside research: TKD, the Scully Food Co-Op, the Anime/Manga club, and tuba.
Education
Research experience
- Led the first-of-kind real-time ML-enabled emission front control experiment at the DIII-D National Fusion Facility.
- Developing foundation models for tokamak plasma data.
- Built TokEye, a fast signal-extraction pipeline for fluctuating tokamak time series.
- Developed reinforcement learning models for visual perception tasks.
- Contributed to Glauber model verification analyses for heavy-ion collisions.
- Improved simulation speed of ORBIT-GPU by 20× on CUDA.
- Investigated 3D visual perception of space and shape; co-authored bioRxiv preprint (2022).
Publications & talks
Acknowledged in 1000 Layer Networks for Self-Supervised RL (NeurIPS 2025 Best Paper Award) for debugging experiments and HPC infrastructure.
Honors
Skills
Programming: Python, PyTorch, MATLAB, C/C++ (CUDA), Bash, Mathematica.
Software & tools: LaTeX, OpenBCI, Microsoft Office Suite, Blender, Photoshop, CAD (SolidWorks, Fusion 360).
Languages: English, Mandarin, Japanese.