Image credit Research & Applied Projects
Projects that move from idea to algorithm, thinking with data and speaking through models.
Master's Project
Green Energy Demand Forecasting under Changing Conditions
Built a daily UK green-energy-demand dataset and compared statistical and neural forecasting models to examine their reliability and failure modes under volatile and changing conditions.
Read MoreCourse Project
Evaluating Dialogue Act Classifiers under Class Imbalance
Compared BiLSTM and CNN-BiLSTM dialogue-act classifiers on the Switchboard corpus, showing how approximately 70% overall accuracy concealed severe minority-class failures.
Read MoreCourse Project
Cost-Aware Search for London Underground Route Planning
Compared BFS, DFS, UCS, and zone-guided heuristic search on London Underground routes, reducing node exploration by approximately 70–85% with heuristic guidance while UCS consistently achieved minimum observed travel cost.
Read MoreCourse Project
Evolutionary Neural Architecture Search for Farsi Vowel Classification
Used a Genetic Algorithm to search feed-forward neural-network architectures for six-class Farsi vowel classification, examining how evolutionary search behaviour and validation performance changed across population sizes.
Read MoreCourse Project
Maximum Likelihood Estimation and Distribution Diagnostics for Dow Jones Returns
Implemented Maximum Likelihood Estimation for Dow Jones daily returns, validating the estimator against SciPy while examining the distinction between correct parameter estimation and an appropriate distributional model.
Read MoreCourse Project
Statistical Robustness under Heavy-Tailed Distributions
Used simulation and distributional analysis to examine when classical LLN and CLT guarantees remain reliable and how heavy-tailed data can cause standard asymptotic reasoning to fail.
Read MoreGraduation Project
Multi-View 3D Vehicle Reconstruction from Video
Developed a multi-stage computer-vision pipeline combining YOLOv5, MiDaS depth estimation, camera calibration, feature detection, and geometric alignment to reconstruct vehicles from multiple video viewpoints.
Read MoreInternship Project
Time Series Forecasting for Siemens Electrical Product Sales
Analysed five years of Siemens electrical-product sales data and developed a SARIMA forecasting workflow to support demand planning and operational decision-making.
Read MoreCourse Project
Functional Enrichment of Estradiol-Responsive Genes in MCF7 Cells
Analysed differentially expressed genes from estradiol-treated MCF7 cells using functional enrichment and multiple-testing correction to identify biological processes associated with hormone-responsive transcription.
Read MoreCourse Project
Unsupervised Clustering of Pan-Cancer Gene Expression
Analysed gene-expression profiles from 801 cancer patients across 20,531 genes using K-Means, PCA, and t-SNE, comparing alternative cluster-selection criteria with known cancer types.
Read More