Projects
Graph Neural Network for Finite Element Analysis
- Built a Graph Neural Network (GNN)-based surrogate model to accelerate parameter search in Finite Element Analysis (FEA).
- Achieved 99.3% accuracy on simulation data, enabling faster exploration of pressure and temperature parameters.
Link Prediction in Citation Networks
- Predicted missing edges in citation networks using Graph Neural Networks (GNNs) and semantic information.
- Ranked 2nd on Kaggle for this project.
- Analyzed the impact of Twitter posts from top 500 accounts on the UK FTSE index using NLP and time series analysis.
Remote Sensing with Deep Learning
- Automated image annotations for UAV-acquired images using UNet and PSPNet models.
- Achieved a Kaggle rank of 11 for this project.
Tabular Data Generation with CTGAN
- Researched and implemented CTGAN for generating synthetic tabular data in the context of Vehicle Telematics.
- Presented findings to stakeholders, enabling the creation of synthetic driving data portfolios.
Covariance Matrix Adaptation (CMA-ES) Algorithm
- Contributed to an open-source implementation of the CMA-ES algorithm, a derivative-free optimization technique.
- Applied the algorithm in an advanced optimization course project.
Data-Driven EEG Band Discovery
- Developed a Python-based strategy for discovering optimal EEG bands using decision trees.
- Outperformed traditional band boundaries by a factor of two, enabling better characterization of brain activity.
