Portfolio

Summary

AI student with a foundational background in Data Science and ML/DL. I enjoy exploring and learning through hands-on projects, and I continuously work on improving my programming, data processing, and model-building skills to prepare for more complex projects in the future.

Technical Skills

Category	Skills
Programming	Python, C++, SQL
Python library	Pandas, Numpy, Matplotlib, Seaborn, Scikit-learn, Opencv, Mediapipe, Pytorch
Tools	Git, GitHub, Jupyter Notebook, VS Code, PyCharm

Education

Artificial Intelligence

FPT UNIVERSITY

Projects

1. Driver Drowsiness Detection System

Project Report (PDF) | Source Code

Developed a real-time Computer Vision system to detect driver fatigue using webcam video, designed with a modular and high-speed processing pipeline powered by MediaPipe.

Key Features

Eye State Classification (Open/Closed): Built a custom lightweight CNN achieving 98.70% accuracy, optimized for low-latency real-time deployment.
Yawn Detection (Hybrid MAR-CNN): Implemented a dual-verification method combining geometric Mouth Aspect Ratio (MAR) + CNN. → Reduced false positives compared to MAR-only approaches.
Real-time System Architecture: Facial landmark detection via MediaPipe, Eye & mouth ROI extraction module, Hybrid classification engine, Alert trigger subsystem.

My Contributions

Implemented data collection, model training, and evaluation steps.
Implemented facial landmark extraction and real-time processing.
Developed custom CNN models for eye state classification and yawn detection.

2. Heart Disease Prediction Using Machine Learning

Project Report (PDF) | Source Code

Applied supervised machine learning models on a clinical dataset (918 samples, 12 features) to predict the presence of heart disease, focusing on minimizing False Negatives (FNs).

Key Features

Comparative Model Evaluation: Trained and evaluated Logistic Regression, XGBoost, and Random Forest classifiers. Random Forest achieved the best results with 88.0% Accuracy and a high 89.0% Recall (crucial for medical diagnosis).
Data Analysis & Feature Engineering: Performed comprehensive Exploratory Data Analysis (EDA) to understand feature distributions and correlations. Identified Oldpeak, MaxHR, and Age as the strongest predictors for the target variable.
Statistical Insight Extraction: Generated heatmaps, pairplots, and count plots to visually confirm the correlation structure. Established that a lower Max Heart Rate (MaxHR) and higher Oldpeak score strongly correlate with increased disease risk.

My Contributions

Led the Exploratory Data Analysis (EDA) and Data Visualization stages.

Contact

Email: pham.phuc.david@gmail.com
GitHub: https://github.com/Pham-Hoang-Phuc