CV

Software RD Intern

Moldex3D

Integrating NVIDIA-Modulus with Moldex3D (CAE Software) and applying Fourier Neural Operator for non-Newtonian fluid thermal analysis and plastic injection.

• High-performance Scientific Computing
• Physics-based Machine Learning
• Computational Fluid Dynamics

Deep Learning Summer Intern (AI/Modulus [NVIDIA - SimNet])

National Center for High-performance Computing, National Applied Research Laboratories

Engaging in research focused on Physics Informed Neuralk Networks (PINNs), including Physics Informed Neural Operator (PINO) and Fourier Neural Operator (FNO).

• RAPIDS
• Apptainer
• High-performance Scientific Computing
• Physics-based Machine Learning
• Computational Fluid Dynamics

Undergraduate Research Assistant

National Central University

Engaging in research focused on physics models in space, machine learning, and their combination.

• Numerical Methods
• Physics-based Machine Learning
• Remote Sensing Analysis

Teaching Assistant

National Central University

Volunteering as a TA for the service-learning course to led our department’s freshman to the campus landfill for recycling in collaboration with the General Affairs Division. I also supervised and advocated for resources recycling within the department building.

Student Representative

National Central University

Holding a Learning Effectiveness Committee to collect student feedback and engage in discussions with relevant faculty and staff members, further strengthen the bond between learners and instructors while organized various events as the designated events holder for the department, including the Learning Effectiveness Committee, Yuri's Night, and parties, among others.

• Valedictorian of College

Master of Science

Imperial College London, London, UK

Computational Science and Engineering

• Software Engineering
• Optimization & Inversion
• Parallel Computing
• Computational Math

Summer Visitor

Stanford University, California, USA

Summer Session

• High Performance Computing
• Data Mining and Analysis
• Stochastic Processes

Bachelor of Science

Natioanl Central University, Taoyuan, Taiwan

Space Science and Engineering

• Numerical Linear Algebra
• Mathematical Modelling
• Plasma Physics
• Remote Sensing Analysis
• Attitude Determination and Control System

Dean's Award of College of Earth Science

College of Earth Science, National Central University

Academic performance in the top 5%, recommended for selection by the department's curriculum committee.

Honor for Academic Excellence

Department of Space Science and Engineering, National Central University

Placed in the top 1% of the department for the semester. (2023 Spring Term)

Best Technical Award

AI Space Challenge, ASEAN

Proposed a particulate matter sensor integrating atmospheric physics parameters and geomagnetic sensing on the International Space Station with Bi-GRU, which was specially awarded by Geo-Insight, for demonstrating the best design and technology among all participating teams (a total of 34 teams) from ASEAN countries.

(Submitted) Regression-based Physics Informed Neural Networks (Reg-PINNs) for Magnetopause Tracking

International Conference on Scientific Computing and Machine Learning 2025 (SCML2025)

Previous research in the scientific field has utilized statistical empirical models and machine learning to address fitting challenges. While empirical models have the advantage of numerical generalization, they often sacrifice accuracy. However, conventional machine learning methods can achieve high precision but may lack the desired generalization. The article introduces a Regression-based Physics Informed Neural Networks (Reg-PINNs), which embeds physics-inspired empirical models into the neural network’s loss function, thereby combining the benefits of generalization and high accuracy. The study validates the proposed method using the magnetopause boundary location as the target and explores the feasibility of methods including Shue et al. [1998], a data overfitting model, a fully-connected networks, Reg-PINNs with Shue’s model, and Reg-PINNs with the overfitting model. Compared to Shue’s model, this technique achieves approximately a 30% reduction in RMSE, presenting a proof-of-concept improved solution for the scientific community.

Automatic Emergency Dust-Free solution on-board International Space Station with Bi-GRU (AED-ISS)

AI Space Challenge

With rising concerns about PM2.5 and PM0.3 particulate matter, which pose threats to both the environment and human health, as well as to instruments on the International Space Station (ISS), our team aims to relate particulate concentrations to factors like magnetic fields, humidity, acceleration, temperature, pressure, and CO2 concentration. Our goal is to develop an early warning system (EWS) using Bi-GRU algorithms to predict particulate levels, providing astronauts with reaction time to protect instruments and increase measurement accuracy. This model could also serve as a remote-sensing smoke alarm prototype.

Utilizing Pop-Up Platform on the CubeSat to Achieve Commercial Activities in the Universe (UCCU)

International Conference on Astronautics and Space Exploration (iCASE)

The mission aims to develop a CubeSat that records videos of messages and objects sent from Earth. These messages can be advertisements, personal messages, or any sponsor-desired objects. The project seeks to showcase the business potential of space advertising and encourage industry-academia collaboration in Taiwan. Key technologies include a pop-up structure, camera design for clear capture of objects and Earth, and radiation-proof OLED displays. An undergraduate team will develop the CubeSat, offering a flexible and affordable space experience. The project also aims to reduce costs in scientific research by integrating commercial objectives.

Pretrained Fourier Neural Operator for Non-Newtonian Fluid Dynamics

Examined the performance of FNO on CAE problems, specifically in the field of non-Newtonian fluid dynamics.

• Fourier Neural Operator
• Numerical Methods
• Computational Fluid Dynamics
• Benchmarking

Regression-based Physics Informed Neural Networks for Magnetopause Tracking

In this study, we propose a Regression-based Physics-Informed Neural Networks (Reg-PINNs) that combines physics-based numerical computation with vanilla machine learning. This new generation of PINNs overcomes the limitations of previous methods restricted to solving ordinary and partial differential equations by incorporating conventional empirical models to aid the convergence and enhance the generalization capability of the neural network. Compared to Shue et al. [1998], our model achieves a reduction of approximately 30% in root mean square error.

• Physics-based Machine Learning
• Magnetopause Modelling
• Benchmarking

Numerical Methods and Physics Informed Neural Networks in Advection Function

Demonstrated that Physics-Informed Neural Networks (PINNs) have a relatively lower likelihood of dissipation occurrence during long-term evolution compared to traditional numerical solution methods such as the Finite Difference Method and Finite Volume Method.

• Physics-based Machine Learning
• Numerical Methods
• Benchmarking

High Performance Computing with NVIDIA-RAPIDS

Employed RAPIDS to replace Scikit-Learn and trained the EMNIST dataset using SVM on a GPU (NVIDIA - RTX5000), resulting in a significantly reduced training time of only 91 seconds.

• RAPIDS-cuML
• RAPIDS-cuDF
• Image Classification
• Benchmarking

Comparing Kalman Filter with Recurrent-based machine learning on Satellite Attitude Determination

his study utilizes attitude data (position and velocity) from the IDEASSat (INSPIRESat-2) satellite and applies the GRU (Gated Recurrent Unit) machine learning algorithm to predict these parameters. The predicted values are compared with those obtained from a Kalman filter, and results indicate that the Bi-GRU algorithm performs slightly better.

• Kalman Filter
• Bi-GRU
• Attitude Determination and Control
• Benchmarking

Automatic Emergency Dust-Free solution on-board International Space Station with Bi-GRU

With rising concerns about PM2.5 and PM0.3 particulate matter, which pose threats to both the environment and human health, as well as to instruments on the International Space Station (ISS), our team aims to relate particulate concentrations to factors like magnetic fields, humidity, acceleration, temperature, pressure, and CO2 concentration. Our goal is to develop an early warning system (EWS) using Bi-GRU algorithms to predict particulate levels, providing astronauts with reaction time to protect instruments and increase measurement accuracy. This model could also serve as a remote-sensing smoke alarm prototype.

• Bi-GRU
• Environmental Monitoring
• Remote Sensing

Benchmarks on ImageNet and MNIST with self-constructed Neural Networks

With the increasing attention on pure Data-Driven Methods, the Fourier Neural Operator (FNO) shows a great promising prospect in applications such as solving partial differential equations and zero-shot super-resolution in addition to the image classification problem. In this work, we thoroughly investigate the mechanism of FNO and independently implement it via Tensorflow. The benchmarks of the standard MNIST image classification problem against FCN, CNN, and ResNet shows a higher accuracy with the price of longer training (and inferencing) time.

• Fourier Neural Operator
• Image Classification
• Benchmarking