The projects below represent my developing research portfolio, spanning machine learning, simulation, natural language processing, and visual analytics. Most were completed as part of my MSc in Computational Social Systems at TU Graz; two are published in IEEE Xplore. Each reflects a genuine engagement with a research question — not just a technical exercise, but an attempt to understand something meaningful through data and computation.
IEEE-Published Research
A Smart Polluted Water Overload Drainage Detection and Alert System
Year: 2021 · Type: Peer-Reviewed Publication · IEEE Xplore
Keywords: IoT · Environmental Monitoring · Sensor Systems · Real-Time Detection · Urban Infrastructure
This project developed an IoT-based system for real-time monitoring of drainage conditions in urban environments. Sensors were used to detect abnormal water levels associated with pollution overload, triggering automated alerts for intervention. The project addressed a practical public infrastructure problem through embedded sensing and data-driven response — demonstrating how computational systems can be designed for socio-environmental applications in resource-constrained contexts.
Publication: https://ieeexplore.ieee.org/document/9447664
A Computer Vision System for the Categorization of Citrus Fruits Using Convolutional Neural Networks
Year: 2021 · Type: Peer-Reviewed Publication · IEEE Xplore
Keywords: Computer Vision · CNN · Image Classification · Agricultural Technology · Deep Learning
This project designed and evaluated a computer vision pipeline using a convolutional neural network (CNN) to classify citrus fruits by category. The system was developed as a proof-of-concept for automated agricultural inspection. The project introduced me to deep learning workflows, model training pipelines, and the challenge of building classification systems that generalize reliably to real-world variation.
Publication: https://ieeexplore.ieee.org/document/9501790
MSc Research Projects — TU Graz
Spatio-Temporal Data Reconstruction (2025)
Keywords: Time Series · Spatial Interpolation · KNN · IDW · Fourier Analysis · Data Imputation
This project addressed the methodological challenge of reconstructing a complete time series for a location whose sensor data was entirely absent. Using a combination of spatial interpolation (K-Nearest Neighbors, Inverse Distance Weighting) and temporal methods (Fourier decomposition, polynomial interpolation), the project produced a reconstructed dataset with improved smoothness and accuracy compared to simpler imputation baselines. The project deepened my understanding of the assumptions embedded in different interpolation strategies and the conditions under which each is appropriate.
Gender Bias in Book Reviews (2024)
Keywords: NLP · Sentiment Analysis · Gender Bias · TF-IDF · Statistical Testing
This project used NLP and statistical methods to examine whether gender bias is detectable in how book reviews are written about male versus female authors. The analysis involved data preprocessing, sentiment scoring using TextBlob, TF-IDF term weighting, and hypothesis testing using t-tests and Mann-Whitney U tests. The project highlighted both the analytical possibilities and the interpretive limits of computational approaches to detecting social bias in language.
Smart Export Gateway — A Digital Platform Concept (2024)
Keywords: Platform Design · Data Integration · AI Matchmaking · SME Internationalization
This project addressed how small and medium-sized enterprises (SMEs) can access global markets without the resources of large corporations. Drawing on open trade, tourism, and economic datasets, the project designed a prototype data-driven platform concept featuring AI-based matchmaking between businesses and international market opportunities. The project engaged with questions of data integration architecture, responsible AI design, and the potential of digital platforms in reducing structural inequalities in global trade access.
Spatial Conflict Modeling — An Agent-Based Approach (2025)
Keywords: Agent-Based Modeling · Mesa (Python) · Simulation · Group Dynamics · Conflict
This project investigated how spatial mobility and the density of shared Points of Interest (POIs) influence conflict dynamics between social groups. Using the Mesa library in Python, I designed a simulation in which agents from different groups moved through a shared space and interacted at POIs. Results suggested that higher POI density reduces inter-group conflict by distributing contact events — a finding consistent with contact theory in social science. The project demonstrated how simulation can operationalize and test theoretical propositions about collective behavior.
Superheroes vs. Villains — Predicting Outcomes from Character Features (2025)
Keywords: Machine Learning · Random Forest · Linear Regression · EDA · Feature Engineering
This data mining project built a prediction pipeline for estimating character win probabilities based on structured attribute data. The workflow included data cleaning, feature encoding, exploratory analysis, and training and evaluation of linear regression and random forest models (RMSE, R²). The project served as a comprehensive exercise in the full machine learning pipeline — with attention to feature selection, model comparison, and performance interpretation.
Global Energy Trade Visualization (2024)
Keywords: Visual Analytics · Interactive Visualization · Global Trade · Energy Systems · Python
This project built an interactive visual analytics tool for exploring complex global energy trade relationships. Using large structured datasets on bilateral energy flows, I processed the data and constructed a dynamic visualization allowing users to explore trade patterns across countries and time periods. The project engaged with core questions in visual analytics: representing high-dimensional relational data, supporting both overview and detail exploration, and understanding how design choices shape user perception.