2022
Internship, Merck
- Developed an AI-based automated PD-L1 CPS scoring model using deep learning for a project closely linked to Merck's KEYTRUDA® (pembrolizumab) anti-PD-1 therapy.
- Collaborated with the clinical team to devise test cases, refining the AI tool's accuracy, and assessed its clinical relevance by benchmarking its performance against pathologist scores using data from clinical studies.
- Utilized Python and PyTorch for the development of the AI tool, focusing on automatic data quality control pipeline, cell segmentation, and predictive modeling.
2020 - Now
Lab technician, MBB lab & CBG center, Tulane University
- Manage and maintain the lab's data collection and analysis system, including organizing, cleaning, and analyzing data to support research projects.
- Maintain the lab's website (MBB lab), including creating and posting content, troubleshooting issues, and ensuring that the site is up-to-date and accessible.
- Communicate with collaborators and invited speakers, manage the scheduling of center meetings, and coordinate with members to ensure that all are informed about project progress and deadlines.
- Train new lab members on laboratory techniques, safety protocols, and software programs.
2018 - Now
Research Assistant, Tulane University
- Developed innovative graph deep learning models to analyze fMRI data and predict phenotypes.
- Integrated multimodal data and examined relationships between different modalities to identify crucial biomarkers.
- Implemented cutting-edge machine learning and deep learning algorithms to ensure high precision in the results.
- Collaborated on neuroimaging and brain function studies with TReNDS Center (GSU/Gatech/Emory), DICoN Lab (Boys Town National Research Hospital), and Mind Research Network, supported by NIH and NSF grants totaling over $2 million.
2017 - 2018
Research Assistant, University of Florida
- Developed a comprehensive data preprocessing pipeline for whole-slide images that includes quality control, segmentation labeling, and data augmentation to ensure high-quality, minimize potential biases in the dataset, and diverse data for subsequent analysis and modeling.
- Designed a rule-based CNN model (Nottingham Histologic Grade) for classifying SCC and ADC breast cancer cells, leveraging TensorFlow and PyTorch frameworks.