Information Technology R D

090******* ********************@*****.*** GITHUB LINKEDIN

Thomas Nguyễn

PROJECTS

Chatbot supports children's health

Developed an application to assist parents in monitoring and understanding their children’s situations effectively.

Processed data by formatting PDFs, documents, and web-scraped content using BeautifulSoup for streamlined analysis.

Designed and implemented both Naive RAG and Advanced RAG models for comparative evaluation, leveraging LLamaIndex. Integrated LLMs with GROD to generate comprehensive and contextually accurate responses.

Enhanced answer precision through CrewAI and advanced prompting techniques. Built the application interface using Streamlit and FastAPI for seamless usability and deployment. Evaluated the model's performance with metrics such as BLEU, precision, and brevity penalty. Source: Code

November 2024 - January 2025

TECHINCAL SKILLS

Programming Languages: Python, Java, C/C++, JavaScript Frameworks: Pytorch, OpenCV, Sklearn, Mediapipe, CrewAI, LLamaIndex, Flask Developer Tools: Git, VSCode, Postman, Jupyter NoteBook, Colab, Kaggle. WORK EXPERIENCES

EDUCATION

Ton Duc Thang university (TDTU) Ho Chi Minh City, Viet Nam Bachelor of Computer Science - GPA: 7.2/10 September 2020 - Present Coursework: Object-Oriented Programming, Data Structure and Algorithms, Database Management Systems, Machine Learning, Computer Vision, Natural Language Processing Honors: Second prize for students researching integration with Computer Vision at Ton Duc Thang University (2021 - 2022)

TMA Solution Lab 6

Developed an AI model with Pytorch to detect sign languages in Australia, with APIs designed for seamless backend integration.

Utilized Mediapipe and OpenCV for data preprocessing, capturing 543 landmark points, and applied Z-score normalization for consistent data scaling. Enhanced model accuracy through K-fold cross-validation, Focal Loss optimization, and Data Augmentation techniques such as Affine transformations. Built a sign language detection model achieving 94% accuracy using CNN1D(for feature extraction) and Transformer(leveraging the Attention mechanism for sequence analysis) architectures. Thoroughly evaluated model performance using metrics like F1-score, Confusion Matrix, Recall, and Precision.

Implemented APIs using Flask and Django frameworks to facilitate integration and ensure efficient connectivity.

AI Engineer Intern August 2024 - November 2024

Ho Chi Minh City, Viet Nam

Entertainment

Play badminton

Swimming

Singing karaoke

Reading books

CERTIFICATIONS

ACTIVITIES

IELTS

Japanese language

September 2024 - September 2026

IDP

Score: 5.5

January 2020

Score: N5

Japansese

Image Classification (Multi-class classification) June 2024 - July 2024 Developed a basic CNN model incorporating Conv2d, MaxPooling, and BatchNorm2d layers to classify four distinct image categories.

Optimized model performance through hyperparameter tuning, enhancing the accuracy of parameter configurations.

Utilized transfer learning with the pre-trained EfficientNetB2 model and conducted a comparative analysis against the basic CNN model to highlight performance differences. Source: Code

Probabilistic weighted frequent itemset mining over uncertain data streams December 2023 - April 2024 Designed a UML diagram to visualize and better understand the workflow of algorithms. Implemented object-oriented programming principles, specifically encapsulation, alongside data structures such as HashMap, LinkedList, and ArrayList for efficient construction. Optimized code complexity, significantly improving execution speed when processing 67,557 rows of data.

Source: Code

Contact this candidate