onyeka muoka

At Royal Free, I currently build ETL workflows in SQL Server and check five years of outpatient activity for data quality. I turn questions from clinicians and managers into precise, defensible analysis while working with sensitive data under strict governance.

In my project work, I developed a diabetes risk classifier in Scikit-learn that achieved 92% precision and cut manual screening time by 40%. I also built a staged student dropout early warning model in TensorFlow/Keras and XGBoost that contributed to a 15% reduction in dropout rates.

For the Bank of England's Prudential Regulation Authority, I built a sentiment and risk indicator tool for quarterly results announcements under NDA, using BERTopic, FinBERT and the Phi 4 language model. I also built time series forecasting pipelines with XGBoost, LSTM and Auto-ARIMA for Nielsen BookScan data.

At Citiscan Diagnostics, I analysed demand data that supported expanding a single ultrasound unit into a full multi modality centre, growing diagnostic volume by 70%. I led a billing system overhaul that cut processing time by 60%; my seven years helping grow a small diagnostic business also gave me a personal connection to the importance of timely support for small enterprises.