Machine Learning Computer Vision

Nguyen Huu Phan

Hanoi • 038******* • ***************@*****.*** • LinkedIn •Github

EDUCATION

Thang Long University

Hanoi

Bachelor of Science in Artificial Intelligence Expected Graduation:06/2026 GPA: 7.7/10

Relevant coursework: Machine learning, Deeplearning, Data base, Math for AI, Computer vision, Natural language processing

Technical Skills

Programming Languages: Python, C++

Frameworks: Pandas, Matplotlib,Numpy, Pytorch, Scikit-learn, Langchain, Transformers, Streamlit PROJECTS

Traffic Law Chatbot (Team size : 3)

● Description: Built an AI-powered chatbot to assist users in understanding Vietnamese traffic laws and answering common legal questions related to traffic regulations.

● Role :

o Used Selenium to scrape data from official government and traffic law websites o Preprocessed and cleaned legal text data to feed into the chatbot system o Built a simple and interactive web interface using Streamlit for real-time user interaction

● Outcome: Successfully handled over 52 common traffic-related questions with approximately 70% accuracy, improving public access to legal information

Human Pose Estimation System

● Description: Built a system to detect and analyze human body postures from images and videos using computer vision techniques.

● Role :

o Extracted 33 3D body landmarks per frame using MediaPipe Pose from real-time or recorded video o Processed and labeled the pose keypoint data into predefined action classes o Designed and trained a Long Short-Term Memory (LSTM) model to classify temporal pose sequences

● Outcome : Trained an LSTM model that achieved 92% accuracy on the validation set Milk Quality Classification

● Description: Developed a machine learning model to classify milk quality (low, medium, high) based on various physicochemical features.

● Role:

o Preprocessed a dataset containing milk chemical properties (pH, temperature, fat, taste, etc.) o Performed feature scaling, visualization, and correlation analysis o Trained and compared multiple classification models: Decision Tree, Random Forest, and K-Nearest Neighbors

● Outcome: Achieved over 98% accuracy using the KNN mode interpretability of decision rules

Contact this candidate