Problem Statement: Extracting meaningful causal relationships from complex, fragmented ICU datasets like MIMIC-III is a major challenge in healthcare analytics.

Solution: Built a scalable causal discovery pipeline that automates preprocessing, inference using Tetrad, and visual analysis of longitudinal EHR data. Enabled clinicians to explore patient trajectories and identify key factors influencing outcomes.

Tech stack: Python, SQL, PostgreSQL, Tetrad, YAML, Pandas, Matplotlib, DNAnexus

Problem Statement: Understanding multimodal travel patterns and behavioral shifts is critical to incentivising rail usage and reducing car miles across Scotland.

Solution: Built a scalable analysis pipeline to quantify the impact of rail disruptions on mobility using ScotRail and road traffic data. Modelled travel behaviour shifts and estimated carbon savings to inform rail-first transport policies.

Tech stack: Python, PostgreSQL, Scikit-learn, GeoPandas, NetworkX, Knowledge Graphs, Matplotlib, Jupyter, Git

Problem Statement: Enhancing the specificity of mRNA drugs by predicting optimal binding regions for the GAG protein, a key step in targeted therapeutic design.

Solution: Identified potential mRNA-GAG binding sites using BLAST and ColabFold, then applied deep learning to predict and quantify binding potential. Modelled interactions using PST-PRNA4 to support in-silico validation for our industrial partner, bYoRNA.

Tech stack: Python, ColabFold, BLAST, PST-PRNA4, Deep Learning (TensorFlow/Keras), Jupyter, Git

Problem Statement: Design a robust and optimised route for the Olympic flame relay that ensures uninterrupted radio coverage across all regions.

Solution: Engineered a geospatial dashboard integrating geocoding, radio coverage simulation, and real-time visualisation. The solution emphasised reproducibility, robustness, and adaptability, ensuring connectivity throughout the relay path.

Tech stack: Python, OpenStreetMap, Geopandas, Folium, Django, PostgreSQL, Leaflet.js, Git