Machine Learning Ai Engineer
Resume:
* ** * - ©VietCV.io
R E S U M E
NGUYEN TUAN
HUNG
INTERN AI ENGINEER
ABOUT ME
EMPLOYEE PAYROLL MANAGEMENT SYSTEM - GROUP PROJECT 01/2024 - 03/2024
FACIAL EXPRESSION RECOGNITION USING DEEP LEARNING- SELF PROJECT
11/2024 - 12/2024
PROJECTS
I am a nal-year Computer Science student at Ho Chi Minh City University of Industry with a strong foundation in arti cial intelligence, machine learning, programming, and web development. Passionate about AI and its practical applications, with hands-on experience from academic projects and self-study. Eager to apply technical skills as an AI Engineer Intern, contribute to real-world solutions, and grow through collaboration and continuous learning in a professional environment. Developed a Java-based application to manage employee payroll and product-based salary calculations.
Implemented MVC architecture, with separate layers for database connection, data processing (DAO), business entities, and graphical user interface (GUI).
Integrated a relational database to store employee information, salary records, and a endance data.
Provided functionalities for salary calculation, employee management, and report generation.
Designed an intuitive graphical user interface (GUI) for ease of use. Link github:
h ps://github.com/TuanHung1810/Employee_Payroll_Management_ System
Developed a system to recognize facial expressions from images. Utilized VGG16 with TensorFlow/Keras for feature extraction. Applied ImageDataGenerator for data augmentation to improve accuracy.
Trained the model using the Face expression recognition dataset. Evaluated the model using metrics such as accuracy and confusion matrix.
Link github:
h ps://github.com/TuanHung1810/Facial_Expression_Recognition
18/10/2002
tuanhung18102002@gmail.
com
497/19 Pham Van Chieu
Street, Ward 15, Go Vap
District, Ho Chi Minh City
h ps://github.com/TuanHu
ng1810
COMPUTER SCIENCE
HO CHI MINH CITY UNIVERSITY OF
INDUSTRY
10/2020 - NOW
EDUCATION
Programming: C/C++, Java,
Python.
Machine Learning: Classi cation,
Regression.
Deep Learning: CNN, RCNN,
LTSM, YOLO, NLP.
Pa ern Recognition: Parametric
(MLE, BE), Non-Parametric
Estimation.
Framework: TensorFlow, Pytorch
Teamwork, Problem solving skill
Computer vision: Classi cation,
Object detection
SKILLS
GPA: 2.7/4.0
2 of 2 - ©VietCV.io
BUILDING AN APPLICATION TO EXTRACT INFORMATION FROM CITIZEN IDENTITY CARDS USING DEEP LEARNING - GROUP PROJECT 12/2024 - 01/2025
BUILDING A FINANCIAL FRAUD DETECTION SYSTEM USING BENEISH M- SCORE AND LSTM - INTERN
05/2025 - 05/2025
INTERN
03/2025 - 05/2025 PHUONG QUAN TRADING SERVICE CO., LTD WORK EXPERIENCES
Develop a system to automatically extract information from Vietnamese citizen identity cards.
Use YOLO (You Only Look Once) for object detection to locate key elds (name, date of birth, ID number, etc.).
Preprocess images to enhance OCR accuracy, including noise reduction and contrast adjustment, then apply OCR (Optical Character Recognition) techniques to extract text from detected elds.
Store extracted data in a structured database for e icient retrieval.
Develop a web-based application for users to upload images and receive extracted information in real-time.
Link github:
h ps://github.com/TuanHung1810/Vietnamese_Citizen_IDCard Develop a system to automatically detect nancial fraud in Vietnamese listed companies.
The system focuses on analyzing nancial statements to compute the Beneish M-Score – a statistical model used to detect earnings manipulation.
Use custom Python scripts and the vnstock library to collect nancial data
Financial reports (Income Statement, Balance Sheet, and Cash Flow Statement) are collected for 100 companies from 2016 to 2024. Data is preprocessed and structured into time series format for analysis.
Apply Beneish M-Score calculation as a traditional nancial fraud detection technique
8 M-score indices( DSRI, GMI, AQI, SGI, DEPI, SGAI, TATA, and LVGI) are used to evaluate the likelihood of earnings manipulation. Train a deep learning model (LSTM) to predict fraud risk Financial time-series data is transformed into sequences and labeled based on M-Score thresholds. Data augmentation techniques (Gaussian noise and SMOTE) are applied to balance the dataset. A Bidirectional LSTM model is trained with accuracy up to 95%.
Developed and improved teamwork and project, problem-solving, time management, critical thinking planning skills through collaborative development and task coordination.
Gained foundational understanding of the stock market and e- commerce platforms.
Strengthened data preprocessing abilities and applied academic knowledge.
ENGLISH
LANGUAGES
Toeic
750
Contact this candidate