Over fifteen years in production and development management across video games and defence, now working as a data analyst. M.Sc. in Data Science, B.Sc. in Statistics.
A turn-based tactics roguelite: isometric grid combat on procedurally generated maps across four biomes, where cover, elevation and initiative decide fights before the dice do. Salvage taken from downed machines unlocks classes and abilities that carry into the next run. Gameplay video, screens and design notes on its own page.
A browser endless runner that ships as a single self-contained HTML file, no build step and no dependencies. Every run reports anonymous telemetry to a Sheets backend, and a static analytics site rebuilds on top of it every six hours, published live alongside the game.
The analysis is built to resist its own convenient conclusions. Survival curves show where runs end, with the hazard estimate cut off once too few runs remain to support it. Every behavioural feature is a per-metre rate or a pure ratio, so nothing scores well merely for lasting longer, and an explicit check reports how much each rate is really just a proxy for run length. Death cause by platform is tested with chi-square, control use with a two-proportion test, and k-means separates playstyles. Each panel writes its own verdict from its own result, gated on significance and a minimum effect size so a trivially small finding is not oversold.
A full-stack analytics report for a fictional action-exploration title, demonstrating applied data science in a live-service gaming context. Covers cohort retention analysis, A/B test evaluation, player segmentation, difficulty curve analysis, and additional behavioural metrics, all built on a synthetic but structurally realistic dataset.
An end-to-end machine learning system for predicting Voice Over localization session costs. Built on a synthetic dataset modelling realistic VO production parameters, an XGBoost regression model achieves 13.9% MAPE (Stage 1) and 23.7% MAPE (Stage 2), against baselines of 19.2% and 28.2% respectively. Includes SHAP-based explainability and a Streamlit interface for interactive quote generation.
An operational analytics dashboard demonstrating how internal form-based feedback can be integrated with vendor performance data to support structured supplier evaluation. Designed around a realistic localization vendor management workflow, combining quantitative KPIs with qualitative scoring from internal stakeholder submissions.
An exploratory analysis of compensation trends across Data Science and adjacent roles, built as part of an M.Sc. programme. Covers salary distributions by job title, experience level, company size, remote ratio, and geography using publicly available survey data. Includes geo-located KPI views and cross-filtering across multiple dimensions.