Khoa Nguyen

A research project to understand existing ML model performance in predicting how T-cells recognize pathogens. I developed a framework that improves current models by generating precise data for training. The framework also acknowledges the existence of cross-reactive T-cells, those providing immunity for humans against multiple pathogens in similar structures.

An automated bidding price module for ad slot auctions written in Python. I took charge of preprocessing the dataset for training, integrating and improving the machine learning model for the bidding price forecasting tasks, and providing complete training and serving architecture for production. The purpose of this module is to help suggest an optimal bidding price for different ad slot auctions while ensuring that the number of ad slots for their creative displays is as high as possible.

A proof of concept (POC) project assessing the capability of federated learning in preserving data privacy while training neural networks. I took charge of performing research and conducting the experiment to ensure that the performance of neural networks is not compromised by exposure of data between organizations.

I was one of the research team members and other members who investigated a visual-based method for automatically determining the number of existing clusters in the dataset. Results show that the approach effectively determines the correct number of clusters and recognizes irregular-shaped clusters during estimation.