Data Scientist · MSc Data Science (with Distinction), University of Sheffield
I'm a data scientist specialising in machine learning, statistical inference and large-scale data processing. My work covers the full analytical lifecycle: auditing data quality, building reproducible pipelines, rigorous model validation, and turning results into clear recommendations for non-technical decision-makers. I've applied this across genomics, aviation, political behaviour, academic peer review, and the music industry.
I'm looking for full-time data scientist or data analyst roles and am open to relocation.
Psoriasis Biomarker Discovery with Interpretable Machine Learning · MSc dissertation
I combined differential expression analysis with a nested cross-validated comparison of XGBoost, Random Forest and SVM classifiers, then used SHAP to explain each model's decisions. This narrowed 21,510 genes down to 14 candidate biomarkers, including six not previously linked to psoriasis.
R limma XGBoost Random Forest SVM SHAP nested cross-validation
Big Data Analysis of ICLR Submissions
This is a distributed analysis of 55,906 submissions to a leading machine-learning conference, written to answer the programme committee's questions about its review process. A cross-validated random forest predicts acceptance from reviewer scores with an AUC of 0.962. I validated the pipeline against reference outputs from two earlier releases of the dataset.
PySpark Spark SQL Spark MLlib Databricks
Predicting Airline Passenger Satisfaction
This is a benchmark of interpretable and high-performance classifiers, with leakage-safe preprocessing, 5-fold cross-validation and hyperparameter tuning. Gradient boosted trees reached a test AUC of 0.97, and the logistic regression's odds ratios showed what drives satisfaction most, led by travel purpose and loyalty-scheme membership.
KNIME logistic regression gradient boosted trees cross-validation
Demographic Drivers of the 2016 EU Referendum Vote
This analysis of the British Social Attitudes Survey uses chi-square tests, multicollinearity diagnostics and binary logistic regression. Educational attainment was the strongest predictor: voters with no qualifications had 5.8 times the odds of voting Leave compared with graduates.
IBM SPSS chi-square tests logistic regression survey data
Metadata and Song Popularity in the Streaming Era
This is a set of five accessible visualisations of 1,377 Billboard Hot 100 songs. I chose each chart type based on research into how people read charts, and designed them for colour-blind and low-vision readers.
R ggplot2 tidyverse data visualisation
| Languages | R, Python, and Spark SQL |
| Machine learning | Gradient boosting (XGBoost), random forests, support vector machines, elastic-net and logistic regression, nested cross-validation, hyperparameter optimisation, SHAP explainability |
| Statistics | Hypothesis testing, generalised linear models, odds ratios, multiple-testing correction, differential expression analysis (limma) |
| Data engineering | PySpark, Databricks, distributed processing of nested and semi-structured data, data-quality auditing |
| Tools | tidyverse, ggplot2, KNIME, IBM SPSS, RStudio, GitHub |
MSc Data Science, University of Sheffield: awarded with Distinction. All modules and the dissertation were completed in 2026, and the degree will be conferred in January 2027.
- Dissertation: Psoriasis biomarker discovery with interpretable machine learning. Awarded 88% (Distinction).
- Distinction-level module results: Big Data Analytics (87%), Data Mining and AI (84%) and Data Analysis (72%).