Machine Learning Information Technology

ABOUT ME

EDUCATION

PROJECT

SKILLS

LE MINH TOAN

Ai Engineer Intern

********@**.***.***.** 081*******

Long Thanh My, District 9, Ho Chi Minh City, Viet Nam I am a fourth-year student at the University of Information Technology (UIT) with a strong passion for Artificial Intelligence. I have worked on multiple projects in Machine Learning, particularly in Computer Vision and Natural Language Processing, as well as deploying some models on IoT devices. Additionally, I have conducted research in Multi task Learning. I also have experience in developing and deploying mobile applications, enabling me to integrate AI solutions into real-world applications. University of Information Technology (UIT) 09/2021 - 06/2025 Faculty of Computer Networks and Communications

GPA: 3.6/4

26/05/2003

Programming Languages: Python

Frameworks & Libraries: PyTorch,

TensorFlow, OpenCV, Hugging Face, Scikit-

learn

MLOps & Tools: Docker, Kubernetes, Jenkins

Machine Learning & Deep Learning: Computer Vision, Natural Language Processing, Multitask Learning,

Model Optimization

OTHER SKILLS

Mobile Development: Android (Kotlin/Java), Flutter Frontend: HTML, CSS, JavaScript

Backend: Flask, FastAPI

Embedded Systems & IoT: Arduino, Raspberry

Version Control: Git

Operating Systems & Networking: Linux, Networking

Database: Firebase, SQL, SQLite

Foreign language: Toeic SW 270

VideoE 2/2025 - PRESENT

Description: A project utilizing the Real-ESRGAN model to enhance individual frames in a video, significantly improving overall video quality. By applying deep learning-based super-resolution techniques, the project restores details, reduces noise, and enhances sharpness, making low- resolution videos appear clearer and more visually appealing. Team size: 1 members

Technologies: Deep Learning, Computer Vision.

Github Linkedin

ACADEMIC RESEARCH

SOFT SKILLS

Role & Contributions:

Preprocessed medical image datasets and structured health-related data for analysis. Designed and implemented a CNN model for pneumonia detection from chest X-ray images. Developed machine learning models to evaluate users' workload and prevent overexertion. Deployed backend services and deep learning models on Kubernetes for scalability and reliability.

Developed and programmed embedded circuits for measuring health indicators. Designed mobile app UI/UX and database.

Reference:

Link Github to model: Link

Link Github to App: Link

Link Demo: Link

Optimizing Multi-Task Learning for Alzheimer’s Stage Classification and MMSE Score Prediction with T-Hybrid Loss

8/2024 - PRESENT

Teamwork &

Collaboration

Description: This study focuses on diagnosing Alzheimer's disease stages and predicting MMSE scores using Multi-Task Learning (MTL). We process 3D brain MRI scans and metadata to train a 3D-ResNet model. A key contribution of this work is T-Hybrid Loss, a novel optimization method combining GradNorm and an enhanced Frank-Wolfe algorithm to dynamically balance task- specific losses by adjusting gradients. This approach improves learning efficiency across tasks and enhances overall performance.

Reference:

Link Github: Link

Time Management

& Prioritization

Research &

Paper Analysis

Adaptability &

Fast Learning

DoctorOn 9/2024 - 01/2025

Description: Developed a health monitoring system consisting of a mobile application, a health prediction model deployed via Flask, and an embedded circuit for measuring heart rate and other vital indicators. The system includes a deep learning model for pneumonia detection from chest X-ray images and machine learning models to assess overexertion during work. Team size: 2 members

Technologies: Deep Learning (CNN), Machine Learning, Computer Vision, Kubernetes, Android, Firebase, Arduino.

Role & Contributions:

Developed a pipeline for video enhancement using the Real-ESRGAN model, implementing an efficient workflow to process and upscale video frames. Experimented with optimization techniques, including knowledge distillation, to create a lightweight model that learns to enhance video quality from Real-ESRGAN while reducing computational cost.

Reference:

Link Github: Link

Video demo: Link video

Link to current research paper: Link

Contact this candidate