Research
Multimodal In-Context Learning for Medical VLMs
● ongoing
Institute of High Performance Computing, A*STAR — Research Intern | Nov 2025 – Present
- Developing a training-free multimodal in-context learning framework enabling medical VLMs to adapt to unseen diseases and domain shifts.
Membrane-Level Cell Segmentation for mIF Imaging
Bioinformatics Institute (BII), A*STAR — Research Intern | Aug 2025 – Mar 2026
- Designed preprocessing pipelines for multiplex immunofluorescence datasets including feature selection and artifact filtering.
- Curated high-quality training samples through manual QC and annotation.
- Fine-tuned pretrained Cellpose-SAM and implemented centroid-based center-cell mask extraction.
DS-CAN: Dual-path Sensor Contrastive Attention Network
Bachelor Thesis | Sep 2024 – Apr 2025
- Proposed a multimodal contrastive learning framework combining accelerometer and gyroscope signals via dual convolutional encoders and attention fusion.
- Introduced temperature-scaled contrastive loss for cross-modal representation learning.
- Achieved 94% accuracy with 1–3% F1 improvement on MEMS gesture datasets.
-
Published in ACM ICDIS 2025 —
DOI
Physics-Informed Neural Networks for Solving PDEs
Data-Driven Intelligent Systems Laboratory — Research Assistant | Sep 2024 – Jul 2025
- Developed Physics-Informed Broad Learning Systems to improve efficiency when solving PDEs.
- Contributed to textbook chapters on Fuzzy BLS and Extreme Fuzzy BLS.
- Manuscript under review at IEEE Transactions on Neural Networks and Learning Systems.
Infectious Disease Simulation with Bayesian Inference
Summer Research | Sep 2024 – Feb 2025
- Implemented epidemic simulations based on SIR models.
- Applied Bayesian inference using MCMC methods including Metropolis–Hastings and Gibbs sampling in R.
- Estimated transmission parameters and quantified uncertainty in epidemic dynamics.
Machine Learning Prediction on Tropical Geometry
University of Birmingham Research Summer School | Jul 2024 – Aug 2024
- Trained machine learning models on 28M quantum period data to infer geometric properties.
- Explored models including MLP, RNN, SVM, LSTM, and GNN.