Data Science, Machine Learning, AI

Nong Duc Thang

AI engineer intern

+ Thu Duc, Ho Chi Minh city # *************@*****.*** 039******* § github.com/Thangnezzz Objective

A 4th-year Data Science student with foundational knowledge in machine learning, deep learning, dataset build- ing, and data processing. Skilled in handling diverse data types, including structured and unstructured. Experi- enced in Python, TensorFlow, PyTorch, and advanced data visualization techniques. Eager to apply analytical and technical skills to solve complex challenges in a dynamic environment. Education

VNUHCM - University of Information Technology (UIT) Major: Data Science

09/2021 – 08/2025

Skills

Technical skills:

Coding: Python, C++, SQL

Data Science & AI: Data crawling, Data processing, Data visualization, Algorithm optimization, AI model development, Machine learning, Deep learning, Image processing, Computer vision, Speech processing, Nat- ural language preprocessing (NLP).

Soft skills:

Languages: Able to read and understand technical documents in English.

Misc: Academic research, Team-work, Leading, Presentation, LaTeX typesetting and publishing. Academic research

DurianLSNet: A Compact and Affordable CNN for High-Precision Durian Leaf Disease Detection:

Description:

08/2024 – 03/2025

Proposed the dataset of 4,437 images with five labels, expertly annotated for high accuracy and reliability.

Introduce DurianLSNet, a lightweight CNN, ensures precise durian leaf disease diagnosis with high efficiency, optimized for mobile and IoT devices.

Our model outperforms traditional CNNs with high accuracy and lower computational cost, making it ideal for real-time durian leaf disease detection on mobile and IoT devices. Technologies: Python, Pytorch, Pytorch-Lightning, Torchvision, Scikit-learn Journal: Computer Networks (Q1) ISSN: 13891286

Process: With Editor

Optimizing ECG Heartbeat Classification with Improved Genetic Algorithm and Stacking Ensembles:

Description:

09/2024 – 03/2025

Evaluating different model architectures on the MIT-BIH Arrhythmia dataset.

Enhancing key Genetic Algorithm steps to improve efficiency and optimize hyperparameters, achieving expected performance with reduced computational cost.

Utilizing Stacking ensemble with Cross-validation to optimize Genetic Algorithm-based models, enhancing performance and reducing overfitting.

Our approach achieves competitive ECG classification on the MIT-BIH dataset without requiring data imbalance techniques.

Technologies: Python, Pytorch, NumPy, Scikit-learn Journal: Circuits, Systems, and Signal Processing (Q2) ISSN: 0278081X Process: With Editor

Nong Duc Thang - Page 1 of 2

Projects

Image Enhancement and Segmentation on Kvasir-SEG Dataset: 02/2023 – 07/2023 Description:

Applied specular removal to enhance the quality of Kvasir-SEG endoscopic images.

Evaluated segmentation models, including SegNet, DuckNet, Unet, UNet 3+, and UNet Xception.

Achieved improved segmentation performance with enhanced images. Technologies: Python, TensorFlow, Scikit-image, NumPy GitHub repository: github.com/Thangnezzz/Enhance-Image-KvasirSEG-Segmentation-Task 2 Aspect Sentiment Quad Predictions for Vietnamese Gameshow Comments on Youtube by Paraphrase Generation:

08/2023 – 02/2024

Description:

Built a Vietnamese social media dataset for aspect-based sentiment quad prediction task.

Applied preprocessing techniques, including HTML removal, acronym normalization, word segmentation, and unnecessary character removal, to enhance data quality.

Utilized the Generation Paraphrase method for task adaptation. Trained and evaluated ViT5 and BARTPho models on the dataset.

Technologies: Python, Pytorch, Pytorch-Lightning, Transformers GitHub repository: github.com/Thangnezzz/ASQP-for-Vietnamese-by-Paraphrase-Generation 2 Fine-Tuning ASR Models: Whisper and PhoWhisper on Vietnamese YouTube Dataset:

02/2023 – 07/2023

Description:

Conducted in-depth research on Automatic Speech Recognition and Whisper architecture.

Explored feature extraction techniques in speech data.

Fine-tuned and evaluated Whisper and PhoWhisper on Vietnamese social media speech data from YouTube, achieving high overall performance.

Technologies: Python, Transformers, NumPy

GitHub repository: github.com/Thangnezzz/Fine-Tuning-ASR-Models-on-Vietnamese-Dataset 2 Hate Speech Detection on Vietnam Social Media Text: 02/2023 – 07/2023 Description:

Collected, labeled, and presented a Vietnamese hate speech detection dataset with 7,300 Facebook com- ments.

Applied preprocessing techniques: lowercasing, removing URLs, special characters, emojis, and normalizing teencode and abbreviations using fine-tuned BARTPho on ViLexNorm.

Evaluated four models: Logistic Regression, SVM, BiLSTM, and ViSoBERT, achieving high overall perfor- mance.

Technologies: Python, TensorFlow, machine learning techniques, Scikit-learn GitHub repository: github.com/Thangnezzz/Vietnamese-Hate-Speech-Detection 2 Nong Duc Thang - Page 2 of 2

Contact this candidate