top of page
My Projects
Welcome to my portfolio. Here you’ll find a selection of my work. Explore my projects to learn more about my work. (Click icons for more projects)
Micro Gas Turbine Electrical Energy Prediction
In this project, I applied regression models to predict the electrical power output of a 3-kilowatt micro gas turbine based on its input control signal over time.
The dataset included over 71,000 time-series measurements, capturing both smooth changes and sudden transitions where the turbine’s output lagged behind the input.
Using regression allowed me to model not just the steady-state relationships, but also the more challenging transitional behaviors where delays made the prediction task less straightforward.
I was motivated by how regression models can go beyond simple curve fitting to capture real-world engineering systems. Since micro turbines play a role in distributed and sustainable energy, being able to accurately forecast their performance with data-driven methods has practical value for improving efficiency, reliability, and energy planning.
The dataset included over 71,000 time-series measurements, capturing both smooth changes and sudden transitions where the turbine’s output lagged behind the input.
Using regression allowed me to model not just the steady-state relationships, but also the more challenging transitional behaviors where delays made the prediction task less straightforward.
I was motivated by how regression models can go beyond simple curve fitting to capture real-world engineering systems. Since micro turbines play a role in distributed and sustainable energy, being able to accurately forecast their performance with data-driven methods has practical value for improving efficiency, reliability, and energy planning.
Classification: A Machine Learning Pipeline for Brain Tumor Analysis
The classification analysis of the brain tumor dataset was designed to predict critical clinical outcomes, particularly the likelihood of a patient receiving radiation treatment, based on a variety of patient characteristics and tumor attributes. The dataset encompassed features such as age, gender, tumor type and size, tumor stage, growth rate, surgical intervention, chemotherapy, MRI results, and family history.
A robust machine learning pipeline was constructed, incorporating tailored preprocessing for both numerical and categorical features—including imputation, scaling, encoding, and feature engineering (interaction terms, polynomial expansion, and binning). Multiple classification algorithms were evaluated, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Classifier (SVC), and K-Nearest Neighbors.
Model performance was assessed using standard classification metrics: Accuracy, Precision, Recall, F1 Score, and ROC AUC. While baseline models showed modest predictive performance, hyperparameter tuning and pipeline optimization slightly improved generalization. Gradient Boosting delivered the most consistent results post-tuning, with modest gains in ROC AUC and F1 scores.
Overall, this classification analysis provides a foundational approach to identifying treatment likelihood patterns and understanding the factors influencing clinical decision-making in brain tumor care. These insights can inform targeted intervention strategies and guide future predictive modeling efforts in oncology-focused datasets
A robust machine learning pipeline was constructed, incorporating tailored preprocessing for both numerical and categorical features—including imputation, scaling, encoding, and feature engineering (interaction terms, polynomial expansion, and binning). Multiple classification algorithms were evaluated, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Classifier (SVC), and K-Nearest Neighbors.
Model performance was assessed using standard classification metrics: Accuracy, Precision, Recall, F1 Score, and ROC AUC. While baseline models showed modest predictive performance, hyperparameter tuning and pipeline optimization slightly improved generalization. Gradient Boosting delivered the most consistent results post-tuning, with modest gains in ROC AUC and F1 scores.
Overall, this classification analysis provides a foundational approach to identifying treatment likelihood patterns and understanding the factors influencing clinical decision-making in brain tumor care. These insights can inform targeted intervention strategies and guide future predictive modeling efforts in oncology-focused datasets
Brain Tumor Regression Pipeline and Analysis
This project presents a comprehensive machine learning pipeline for the regression analysis of a brain tumor dataset, with the objective of understanding factors influencing patient survival and predicting clinical outcomes.
This focuses on predicting patient survival rates based on a comprehensive set of clinical and demographic variables. Utilizing multiple regression algorithms, including Linear Regression, Ridge, Lasso, Decision Tree, Random Forest, Gradient Boosting, Support Vector Regression (SVR), and K-Nearest Neighbors—this analysis aims to quantify the impact of features such as age, tumor size, tumor growth rate, stage, treatment modalities, and patient history on survival outcomes.
Advanced preprocessing techniques, including polynomial feature expansion, interaction terms, and binning of continuous variables, were integrated into a streamlined modeling pipeline. Each model was evaluated using standard performance metrics: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²) to assess predictive accuracy and generalizability.
The regression results revealed that ensemble models such as Random Forest and Gradient Boosting provided near-perfect fits (R² ≈ 1.00), suggesting strong predictive capability when capturing complex nonlinear relationships in the data. Moreover, visualization techniques like KDE plots and distribution heatmaps were employed to interpret the relationship between tumor progression and survival outcomes.
Overall, the regression analysis offers valuable insights into prognosis and can potentially support personalized treatment planning by identifying key predictors of long-term survival in brain tumor patients.
This focuses on predicting patient survival rates based on a comprehensive set of clinical and demographic variables. Utilizing multiple regression algorithms, including Linear Regression, Ridge, Lasso, Decision Tree, Random Forest, Gradient Boosting, Support Vector Regression (SVR), and K-Nearest Neighbors—this analysis aims to quantify the impact of features such as age, tumor size, tumor growth rate, stage, treatment modalities, and patient history on survival outcomes.
Advanced preprocessing techniques, including polynomial feature expansion, interaction terms, and binning of continuous variables, were integrated into a streamlined modeling pipeline. Each model was evaluated using standard performance metrics: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²) to assess predictive accuracy and generalizability.
The regression results revealed that ensemble models such as Random Forest and Gradient Boosting provided near-perfect fits (R² ≈ 1.00), suggesting strong predictive capability when capturing complex nonlinear relationships in the data. Moreover, visualization techniques like KDE plots and distribution heatmaps were employed to interpret the relationship between tumor progression and survival outcomes.
Overall, the regression analysis offers valuable insights into prognosis and can potentially support personalized treatment planning by identifying key predictors of long-term survival in brain tumor patients.
Operation Defend the North Vancouver 2025: White Paper
A fictitious company ODTN Telecoms, recently executed a swift and coordinated incident response operation following a targeted cyber-attack on its internal network.
The breach was first detected through abnormal login activity and unauthorized access attempts. The security team immediately activated the incident response plan, isolating affected systems to prevent lateral movement. Forensic analysis confirmed the presence of malware designed to exfiltrate customer data.
Containment procedures included network segmentation and credential resets. The team collaborated with external cybersecurity experts to assess the scope and neutralize the threat.
Communication protocols ensured stakeholders and regulators were informed in real-time. Within 48 hours, systems were restored with enhanced monitoring in place.
A full post-incident review identified gaps and recommended policy updates. ODTN reaffirmed its commitment to data security and operational resilience.
The breach was first detected through abnormal login activity and unauthorized access attempts. The security team immediately activated the incident response plan, isolating affected systems to prevent lateral movement. Forensic analysis confirmed the presence of malware designed to exfiltrate customer data.
Containment procedures included network segmentation and credential resets. The team collaborated with external cybersecurity experts to assess the scope and neutralize the threat.
Communication protocols ensured stakeholders and regulators were informed in real-time. Within 48 hours, systems were restored with enhanced monitoring in place.
A full post-incident review identified gaps and recommended policy updates. ODTN reaffirmed its commitment to data security and operational resilience.
.jpeg){kind=small}
bottom of page