Research Projects
My research interest is in optimizing and implementing machine learning algorithms to facilitate research processes in different sectors.
Anchor Alignment Adaptive Weighting for Robust Domain Generalization
- Developed novel algorithm leveraging NLP anchors to enhance domain generalization and noise robustness.
- Outperforms ERM, IRM, Mixup, GroupDRO, and VREx by 6-13% across five benchmarks under noisy labels.
- This research is under review at Transactions on Machine Learning Research (TMLR).
- [View more] [Paper] [PDF] [Poster]
DeformX: A Versatile Co-Simulation Framework for Deformable Linear Objects
- Developed a co-simulation framework integrating Cosserat rod physics with NVIDIA Isaac Sim for realistic DLO simulation.
- Generated WireSeg-32k dataset (32,000 synthetic images) achieving 10.2% mAP@75 improvement in wire segmentation.
- Demonstrated sim-to-real transfer with 6.6 cm mean error on UR5e rope-swinging task.
- This research is submitted to IROS 2025.
- [View more] [Paper]
RNN Optimization
- Optimizing the recurrent neural network architecture through integration with diffusion probabilistic model.
- [Code]
Conformal Aerosol Jet Additive Manufacturing
- Introduce an approach to designing integrated circuit that is conformal to provided geometry.
- Built multi-axis conformal printing with robot arm to fabricate functional circuits.
- This research is published at CASE 2025.
- [View more] [Video] [Video Abstract]
Vision-based Close-loop FDM Printing for Fabricating Airtight Structures
- Introduce a system to greatly improve print quality of FDM printer with computer vision.
- [View more] [Paper] [Video]
ALS Genetic Motor Neuron Susceptibility
- Predicting genetic causes for motor neuron to be more vulnerable in ALS based on known disease genes and their pathways using machine learning.
Smart Laser Lithotripsy System
- Develop and incorporate machine learning models to create a fail-safe system for the medical device.
- Implemented data logger processor program to filter and visualize data from thermocouples.
- [Website]
Multi-Modal Analysis of Teacher Negativity in Classroom Observation Videos
- Proposed and implemented an automatic and semi-automatic negativity filter technique that can help teachers and learning science researchers find critical moments within classrooms, which is essential to improve teaching quality.
- [Paper]
Speaker Recognition System
- Implemented and modified SpeechBrain into a real-time speech activity and speaker recognition system based on PyTorch with user interface. Enabled a client-server data transfer pipeline to accelerate the speech processing speed.
- [Video] [Code]
Class Projects
Elderly-Care Connect Phone Software
[Video]
Activation Function Design for Fake Face Detector
[Code]
Emotion Classification Machine Learning Model
[Code]
