PhD Student

Mechanical and AI Lab • Department of Mechanical Engieering •
Carnegie Mellon University

Research Objective:

Develop deep generative models to improve the efficiency and safety of computational tools

Expect to graduate on 2024 Summer • Open to work

Google Scholar

LinkedIn

GitHub

Selected Projects

• A physics-informed diffusion model for high-fidelity flow field reconstruction (paper | code)

A denoising diffusion probablistic model for robust super-resolution of turbulent flow data. Currently in the process of being incoprated to modulus.

• Scalable Transformer for PDE Surrogate Modeling* (paper | code)

A computationally more efficient neural PDE solver using multi-dimensional factorized attention.
(* second author, paper accepted for NeurIPS 2023 poster session.)

• Spatial Completion of Turbulent Data Using Deep Generative Models* (paper | code)

A study on reconstructing the missing region in turbulent flow data with vector-quantized generative adversarial network model. (* code will be available after paper published)

• Talking-head Video Synthesis Using a Face-transformation generative model* (paper | code)

A study to investigate STEM learners’ ability to decipher AI-generated video created by a face-swapping generative model. (* The face-swapping model is adopted from the work by A. Siarohin et al. The driving video below was made by the courtesy of Mitchell Fogelson.)

• 3d design using generative adversarial networks and physics-based validation* (paper | code)

A deep generative model to synthesize 3D mesh objects for evaluating a design cycle consisting of synthesis and physics-based validation. (* The code is currently being updated to support newer version of PyTorch and CUDA toolkit. Will become available soon. )

• Generative Adversarial Attacks Against Intrusion Detection Systems Using Active Learning (paper | code)

A generative adversarial active Learning method to model query-efficient attacks against network intrusion detection systems for safefy evaluation.