cv

Bioinformatics Data Analyst

Huntsman Cancer Institute

Developed a full-stack web application for visualizing and analyzing single-cell RNA sequencing data (scRNA-seq) using JavaScript-based technologies (React JS, Node JS, HDF5, Git, Yarn, Linux, Docker). Analyzed single-cell data to identify significant of genes expression differences across cell types based on responses and time point; performed two-way ANOVA to find the interactive term. Implementing semi-supervised learning (Leiden) to analyze scRNA-seq data to track transcriptional state shifts, expansions, emerging, and disappearing states in patient tumor samples over time and response to therapy.

• web development, single cell data analyst

Software Engineer

Rakuten Mobile Corp

Designed and developed an auto-monitoring dashboard for real-time Linux server health management,particularly for 4G and 5G Radio Access Network (RAN) data analysis, to improve monitoring, enhance network performance, and facilitate informed decision-making. This includes: Designed an automated system to log in, collect health data, and capture resources, users, and connectivity every 4 hours. Analyzed and visualized data on a Streamlit dashboard. Summarized weekly KPIs for cluster status and connectivity, and identified anomalies for troubleshooting.

• Dashboard report
• data visualization

Research Assistant

Developed computer vision-based student engagement detection in online classes. Performed face detection using MTCNN and emotion detection using Mini–Xception. Analyzed the final results and created a dashboard called MOEMO to visualized the result in more interactive way. Generated after-class report that report comprehensively details students’ affective states, concentration, engagement, and intervention time for each student.

Master

Hosei University

Science and Engineering

Bachelor

VNU-HCM University of Information Technology

Computer Science

Exploring the Use of CLIP Model for Images Recommendation in Noun Memorization using Various Learning Context

Bulletin of Research Center for Computing and Multimedia Studies, Hosei University

this paper developed a web-based system for noun learning that creates learning materials using the CLIP-model recommended images and translation data from translation API.

A Real-Time Learning Analytics Dashboard for Automatic Detection of Online Learners’ Affective States

Sensors 23 (9), 2023

paper presents a novel learning analytics system called MOEMO (Motion and Emotion) that could measure online learners’ affective states of engagement and concentration using emotion data.

Frame Adaptive Rate Control Scheme for Video Compressive Sensing

International Conference on Image Analysis and Processing (ICIAP) 2022

This paper presents a frame adaptive rate control scheme for measurement coding improving the video quality up to 1.56 dB PSNR and reduces the processing time up to 53% compared to the state-of-the-art

Can Sakai Log Data Improve Learning Analytics? Findings from a Preliminary Survey

33rd Education and Learning Support Information System Research Presentation, 2021

This paper reviewed LAK, EDM, and ACM proceedings and associate journals to shed light on Sakai LMS's potential as it offers flexible tools for teaching, learning, and dynamic collaboration. In comparison with other LMSs, Sakai was less explored by the learning analytics community

Briefing and Geo-visualizing on International Practices of Learning Analytics in Higher Education

The 21st IEEE International Conference on Advanced Learning Technologies (ICALT) 2021

This paper briefed on the selected cases and practices carried out in the last decade where learning analytics, educational data mining, and statistical methods are used in developing predictive models, smart learning tools, and learning analytics tools. In addition, the geo-visualization method is adopted to visualize some promising outcomes of learning analytics between the years 2017 to 2020 in the educational contexts of Asia, North America, South America, the UK, and Europe.

Students’ Emotion extraction and visualization for engagement detection in online learning

25th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES) 2021

This paper focuses on analyzing online lecture videos to detect students’ engagement without relying on learning management systems produced data.

Bi-directional intra prediction-based measurement coding for compressive sensing images

IEEE 22nd International Workshop on Multimedia Signal Processing (MMSP) 2020

This work proposes a bi-directional intra prediction-based measurement coding algorithm for compressive sensing images improving 0.01 - 0.02 dB PSNR improvement and the birate reductions of on average 19%, up to 36% compared to the state-of-the-art

Kidney Pathology Image segmentation

Performing kidney pathology image segmentation to segment chronic kidney disease (CKD) glands in patch images by training and testing multiple baseline CNN models (UNet, Attention UNet, RegNet, DynUNet) and Transformer models (UNetR, Swin UNetR) using the MONAI (Medical Open Network for AI) network architecture. Evaluate the accuracy of network combinations (CNN and Transformer) against the baseline models.

Image Search In Context with Contrastive Language–Image Features

Developed a web-based noun learning system utilizing the CLIP (Contrastive Language–Image Pre-training) model for image recommendations, Azure speech to text, translation APIs for user voice input and translation, leveraging Streamlit for a user interface.

Surveillance network locates individuals using computer vision

Developed a real-time system that can search a person by their face/ attribute in multiple cameras. Collected and preprocessed the real data from the university surveillance cameras. Implemented face detection using MTCNN, and pedestrian detection using YOLO, face cluster by K-mean.

Chat Agent

Using pre-trained Large Language Models from Hugging Face API to create a chat conversation. The chat agent tokenizes the text input, generate the embedding output, then, decode the embedding vector to create the text output.

Customer Intent Classification

Train LSTM (using Tensorflow), BERT (using PyTorch) models to classify the customer's intentions on the BANKING77 dataset (13,083 customer questions, 77 classes of intention). Achieve 85% and 91.46% sequentially accuracy on test set.

Encoder Enhancement For Compressive Sensing

Designed a new encoder method for compressive sensing images by exploring the Walsh Hadamard matrix structure, resulting in a 19% reduction in file size and enhancing compression image quality. Proposed a Triangle Quantization method for compression rate control that allows the system to adjust the image quality upon the internet bandwidth.

Self-Drive Taxi Simulation

Train and test an agent to pick up passengers and drop them at their destination in the fastest way possible using Reinforcement Learning (Q-learning/SARSA). Simulate the environment using Gymnasium library.