My Projects
A collection of my recent work and personal projects. Each project represents unique challenges I've tackled and skills I've developed.
Application of Machine Learning Methods in Financial Portfolio Optimization
Developed a financial portfolio optimization model that outperformed the S&P 500 index by 39% in cumulative returns, using Cat Boost gradient boosting algorithms. Implemented multiple asset allocation strategies including equal-weight, inverse volatility, and ML probability-based weighting. Achieved ROC-AUC score of 0.586, F1 score of 0.638, and Jensen-alpha over 17% across portfolios. Utilized Python (pandas, numpy), scikit-learn, and financial analysis libraries for feature engineering, model development, and backtesting.
Homelab Server
I snagged a budget-friendly HP Elitedesk for just €110 with an i7 8700 processor and 16GB RAM, then upgraded it to 48GB RAM with a 2TB HDD and 500GB SSD. I've transformed it into a versatile homelab server running Proxmox and Docker, hosting multiple virtual machines and containers. My setup now runs several useful services including my personal website, Pi-hole for ad-blocking, a Wireguard VPN server for secure access, Immich for keeping my photos safely backed up, and TrueNAS for all my file storage needs.
A Barber E-Commerce Website - Fullstack Application
Developed a Java Spring Boot e-commerce backend with user management, product catalog, and order processing capabilities. Implemented secure authentication using Spring Security with role-based access control. Created RESTful APIs for product management with filtering, searching, and pagination. Built a complete order workflow system with status tracking and inventory management. The frontend is built with HTML/CSS/JS, and provides both customer shopping experience and admin dashboard functionality.
Electricity Usage Forecasting
Implemented a machine learning project to predict household electricity usage using time series data. Created multiple imputation methods for missing values, including mean, KNN, and MICE, with MICE providing best results. Developed feature engineering pipeline with lag features for time series prediction. Evaluated multiple regression models (Linear, Random Forest, XGBoost, LightGBM) with LightGBM performing best. Used Optuna for hyperparameter tuning to optimize model performance. Achieved 92% R-squared and RMSE of 0.178, demonstrating strong predictive accuracy. Generated confidence intervals for predictions and visualized model performance with error analysis.
WeatherNow App - iOS Application
A simple iOS weather application that provides real-time weather updates and forecasts. Built using Swift and SwiftUI, the app fetches data from the OpenWeather API and displays it in a user-friendly interface. Users can view current weather conditions for any location.
Enhanced Hybrid Keyer 3.0
A professional-grade chroma key shader for OBS Studio using the obs-shaderfilter plugin. This shader features advanced keying algorithms, precise edge controls, and specialized tools for handling difficult keying situations like hair edges and uneven lighting.
Personal Website
A modern portfolio website built with Astro and TailwindCSS. Features a responsive design with a beautiful animated gradient background, interactive hover cards, a book review section showcasing books I've read, and a blog where I share my thoughts on technology and finance.