Human Centered Computing Lab, University of Michigan

Undergraduate Research Assistant

• Generative Tutorial (co-first author; with Profs. Anhong Guo and Xu Wang) — co-developed a conceptual framework and AR system for live contextualized visual instruction that grounds physical-task goals in users' workspaces and proactively generates workspace-specific goal images and demonstration videos by propagating observed and predicted visual states across task dependencies; co-led a formative evaluation of 176 generated artifacts across 15 tasks and a 24-participant comparative study, showing improved task quality, perceived workspace correspondence, and shorter step-confirmation intervals over pre-authored guidance.
• MoiréDeform (co-first author, ACM MobiCom '26 Demo; with Profs. Alanson Sample and Ke Sun) — developed a low-cost camera-based sensing approach that exploits naturally occurring woven textures as virtual moiré layers to amplify subtle surface deformation, enabling continuous body-lean interaction and respiratory waveform/rate sensing on an unmodified mesh chair without instrumenting the user or surface; reduced respiratory-rate MAE from 2.07 to 0.71 breaths/min relative to physical-marker tracking.
• RespiraFrame (with Prof. Junyi Zhu) — contributed to the experimental design and evaluation of an eyeglass-based multimodal respiratory sensing system combining bone-conduction acoustics and electrical impedance tomography (EIT); designed evaluation protocols spanning respiratory behaviors, motion and environmental interference, a three-day semi-wild study, and repeated wear over a four-week longitudinal study.
• VeilLens (co-first author; with Prof. Ke Sun) — developed an agentic visual privacy framework that autonomously discovers, localizes, and selectively edits weak compositional cues exploited by VLMs, supporting granular user control and iterative verification of residual privacy leakage across identity, geolocation, affiliation, and social-relationship risks.

Generative AI for Music and Audio, University of Michigan

Research Assistant, advised by Prof. Herman Dong

Investigating real-time human–GenAI co-creation for controllable music generation, focusing on interaction mechanisms that preserve user agency, expressive control, and iterative creative feedback; prototyping gesture-driven interaction techniques that connect embodied user input to real-time musical control and generative model behavior.

• SuperConductor — conducting an AI ensemble with the hands; next, the MagicBaton as its instrument.
• Strokes of Sonic Tradition — AI that reads jianzipu guqin tablature (in preparation).

Energy Distance Dimension Reduction, University of Michigan

Selected Researcher, Undergraduate Research Program in Statistics (URPS), advised by Prof. Kerby Shedden

Developed engression and sufficient dimension reduction methods for multilevel, dyadic, and longitudinal wearable data from bone marrow transplant patients, addressing non-i.i.d. structure not handled by existing methodology; implemented SDR algorithms in JAX with GPU acceleration, achieving approximately 7,000× speedup over the original method for U-statistic computation on large sensor data series.

Lab of Atmospheric and Earth System Sciences, Nanjing University

Summer Research Intern, advised by Prof. Minghuai Wang

Studied statistical machine learning methodologies for integrating real-world sensor observations with simulated atmospheric data, enabling inference and prediction of aerosol–cloud interactions across larger spatial and temporal scales.

E2-E Lab, University of Illinois Urbana-Champaign

Undergraduate Research Assistant

Conducted airflow experiments and processed PIV/camera data using Python, OpenCV, and OpenPIV for flow-field reconstruction and particle tracking; performed CFD simulations in ANSYS Fluent to analyze particle–airflow interactions in complex 3D geometries.