Analytical. Adaptable.
Always Learning.
I turn data into stories, models into strategy, and insight into action.
Experience
Projects
Multi-Channel Attribution
Analysed 3.3M contacts and 240K orders to compare catalogue vs. email campaign effectiveness with PostgreSQL. Computed attribution models (first-, last-, linear-click), revealing catalogue's six times higher conversion. Segmented customers into 125 RFM cells, estimated CLV, and recommended targeting strategies based on breakeven cost and ROI.
Portfolio A/B Testing & Analytics.
Ran A/B tests with VWO to measure impact of clickbait vs professional phrasing on my portfolio website's “Download CV” CTA; the professional version saw a statistically significant 70.3% higher CTR. Drove traffic via social media, Selenium bots, and a targeted Google Ads campaign; tracked conversions and bounce rate using event-based analytics.
Text Analytics Project on Song Lyrics
Modelled 40,000+ lyrics through LASSO regression with n-grams and FastText embeddings to explore song popularity drivers. Discovered eight genre-level trends in explicitness and thematic content. Leveraged PCA-filtering and cosine similarity on word vectors to quantify semantic shifts between artists' first albums and subsequent works.
RecSys Purchase Prediction
Built a two-stage pipeline on 33M clicks and 1.1M purchases to predict purchasing sessions with XGBoost and forecast items bought via heuristic ranking. Engineered 21 features with temporal encodings and rarity filters. Tuned thresholds to optimise a Jaccard-based score and used SHAP to interpret key drivers like session duration and item popularity.
Beat the Bookies
Engineered a Premier League match predictor using 18 features across five seasons (older data degraded performance), achieving 51% validation and ~40% test accuracy with CatBoost. Feature-engineered Pi-rating variants with tuned hyperparameters and selected key variables using SHAP. Explored pre-game sentiment as a predictive signal using LLMs and Nitter API.
RIE Cleaning of Lagged Covariance Matrices
Developed a spectral cleaning algorithm using Rotationally Invariant Estimators, shrinking noise in high-dimensional lagged covariance matrices. Achieved 100% noise reduction in AR(0) and ~20% signal-to-noise improvement in AR(1), with more gains as dimensionality increased. Applicable to portfolio risk modelling, time series forecasting, and signal processing.
Logistics and Supply Chain Optimisation
Simulated a multi-region supply chain implementing hybrid demand forecasting (SARIMA, Croston's method) and cost-benefit analysis to guide factory and warehouse expansion. Applied adjusted reorder points and Silver-Meal heuristics to manage seasonal inventory, securing 2nd place in Imperial's cohort.
Energy Demand Forecasting
Developed and tuned XGBoost model to forecast UK electricity demand, engineering features like Heating Degree Days, holiday dummies, and temporal lags. Achieved 96.3% R²; generalises well with low residual autocorrelation.
SvelteKit and TailwindCSS Portfolio
Developed a responsive portfolio with SvelteKit and TailwindCSS featuring a comprehensive design system using CSS variables for consistent styling. Implemented accessible components with ARIA attributes, smooth Svelte transitions, SEO optimisation, and Google Analytics integration while ensuring performance through Svelte's compiler design.
Education
About Me
I'm Bryan, a data analyst with a passion for transforming complex data into actionable insights. My expertise lies in leveraging advanced analytical techniques and data visualisation tools to uncover trends, patterns, and opportunities that drive strategic decision-making.
My professional journey encompasses equity analytics to complaints data visualisation. This diverse experience has honed my ability to navigate through different data paradigms, allowing me to adapt and apply my skills across various domains. I thrive on challenges that require a blend of technical acumen and creative thinking, and I'm always eager to learn and grow in this ever-evolving field.
I believe great work happens through collaboration. I genuinely enjoy teaming up with people across different departments to solve tricky problems. Whether I'm helping predict trends, improve workflows, or enhance user experiences, I combine detail-oriented analysis with creative thinking to deliver solutions that actually make a difference.