Data analyst with a background in construction HSE, applying machine learning to environmental, safety and health data.
London, UK · MSc Business & Data Analytics (Distinction), Ravensbourne University London
Before moving into data science I was Assistant Director, HSE&Q at China Railway Group Limited (CREC) on a major infrastructure construction project (2022–2025). I ran site inspections and audits, reported incidents and prepared HSE reports for senior management, and saw first-hand how environmental monitoring and safety decisions are made in practice.
That experience shaped my MSc dissertation, which asks whether a breach of an environmental standard can be predicted before the sample is taken, so that monitoring goes where it is needed most. I am now developing it into a conference paper. Across my work I try to make results that hold up: I audit for data leakage, compare every model with simple baselines, validate in time order and explain what drives each prediction.
I currently work as a Fulfillment and Data Associate at 01Health in London, and I am looking for a funded PhD in environmental data science, construction and HSE safety analytics, health data science, or business analytics and operational research.
Corridor Environmental Compliance · MSc dissertation
Predicts whether an air, water or noise reading will breach its national standard before the sample is taken, using 2,777 readings from 38 sites over three years on an anonymised infrastructure corridor.
- Found a target-leakage bug that had inflated earlier results, and rebuilt the features from past readings only
- Nested time-series cross-validation: the model beats a parameter-only baseline in every domain (noise F1 0.549 against 0.255)
- Showed that monitoring effort and risk are misaligned: in every zone, 39–56% of readings go to low-risk parameters
- Ten-tab Streamlit dashboard for prioritising monitoring by parameter and site
Python scikit-learn XGBoost SHAP imbalanced-learn Streamlit Plotly
Amazon Review Sentiment · MSc AI & Machine Learning module
A Bidirectional LSTM that classifies product reviews as positive or negative, tested against a TF-IDF + Random Forest baseline on the same 10,000 held-out reviews.
- 90.0% accuracy (ROC AUC 0.966) against 86.8% (ROC AUC 0.942) for the baseline
- Error analysis shows the whole gain comes from positive reviews, and that negation is still a weak spot
- Found and fixed a train/test mismatch that had left the original coursework baseline at chance level
Python TensorFlow / Keras scikit-learn NLP
Models that help decide where limited inspection, monitoring or healthcare resources should go:
- Environmental data science: predicting compliance risk and prioritising environmental monitoring
- Construction and HSE safety analytics: learning from inspection and incident records
- Health data science: turning routinely collected data into decisions
- Business analytics and operational research: decision support when resources are limited
| Area | Tools |
|---|---|
| Programming and data | Python (pandas, NumPy, SciPy), SQL |
| Machine learning | scikit-learn, XGBoost, TensorFlow / Keras, imbalanced-learn, SHAP |
| Visualisation and BI | Power BI, Tableau, Excel, Streamlit, Plotly, Matplotlib, seaborn |
| Methods | Time-series cross-validation, leakage auditing, class imbalance, clustering, NLP, CRISP-DM |
| Workflow | Git and GitHub, Jupyter, Google Colab |
- Fulfillment and Data Associate, 01Health, London · 2026–present
- Assistant Director, HSE&Q, China Railway Group Limited (CREC) · 2022–2025
Site inspections and audits, incident reporting, HSE reporting to senior management, team supervision and payroll for 100–500 staff - MSc Business & Data Analytics, Distinction · Ravensbourne University London · 2026
- BBA in Finance · University of Information Technology & Sciences
- Certifications: IOSH Managing Safely · OSHA 30-Hour General Industry Safety