Selected work

Eleven projects, each with a full write-up.

Data platforms, pipelines, applied ML and BI work.

2026 · Force24 · Analytics Engineer, Data Layer

Account Intelligence Platform

A greenfield Account Intelligence platform built over 16 weeks at Force24 for CSMs, accounts, and stakeholders, shipped with the engineering team into a live production environment. I owned the data layer end to end, integrated it into the FastAPI service through endpoint changes and Redis caching, and shipped features and enhancements across the Angular frontend. I built it on the principle that a dashboard should display data and also lead the user to a specific action. Confidential, sanitised case study.

Dagsterdbt

PostgreSQL

Python

Redis

Angular

Read the case study →

2026 · MSc dissertation · Data platform · MCP

Agentic ELT Data Platform for Customer Intelligence

Agentic ELT Data Platform thumbnail showing layered data flow with MCP node

MSc dissertation in a live B2B SaaS environment under NDA. End-to-end JSONB-first ELT platform and three-model churn intelligence stack (survival, XGBoost via PostgresML, and DR-Learner causal inference) surfaced via FastAPI, Angular, and an MCP endpoint for agentic LLM access. Over 1 million records ingested across multiple vendor APIs, 48 dbt models.

Python

PostgreSQLdbt DagsterPostgresMLXGBoostMCP

Read the case study → View on GitHub →

2025 · Side project · Healthcare ML

Pharmaceutical Side Effect Classification

Production-grade Python package classifying free-text adverse-event descriptions into a MedDRA-style taxonomy of ten clinical categories across 11,825 marketed medicines. Single sklearn Pipeline with ColumnTransformer for TF-IDF text features and ordinal manufacturer encoding, joblib-serialised end to end. YAML config loaded into frozen dataclasses, CLI entry points, pytest fixtures, GitHub Actions CI matrix on Python 3.10, 3.11, 3.12.

Python

scikit-learn

pandas

pytest

GitHub Actions

Read the case study → View on GitHub →

2023 · MSc Thesis · Sports analytics

Big Data Player Scouting

Ranking and recommending football talent across five top European leagues using event data and PlayeRank metrics. Hypothesis-driven research piece exploring how data analytics can support coaches and scouts in player recruitment. Published thesis, public codebase.

Python

PySpark

Jupyter PlayeRankUEFA event data

Read the case study → View on GitHub →

2026 · Side project · NLP and media analysis

Conflict Sentiment Analysis

A comparative methodology study quantifying how 8,158 English-language news articles framed the Russia and Ukraine conflict across 68 publishers and 18 countries. Three sentiment engines (TextBlob, VADER, CardiffNLP RoBERTa) and a five-topic LDA over the same corpus, behind an abstract SentimentEngine base class with a registry pattern. Transformer mocked in CI so tests do not download gigabytes of weights.

Python

HuggingFace NLTKgensim LDA

GitHub Actions

Read the case study → View on GitHub →

2023 · Side project · Gaming sector

Collaborative Filtering Recommender on Spark

Steam Games Recommender thumbnail with three-node network graph

PySpark + ALS recommender on the Steam 200k implicit-feedback dataset. Distributed training, hyperparameter tuning via Grid Search with CrossValidator, full experiment tracking and model logging in MLflow. Built and tested on Databricks Community Edition.

PySparkSpark MLlibALS

MLflow

Databricks

Python

Read the case study → View on GitHub →

2025 · Side project · Finance, time series

Equity Forecasting

Reproducible R analysis package for daily equity closing-price forecasting. ARIMA, ETS, and a naive baseline behind a MODEL_REGISTRY pattern. Combined ADF + KPSS stationarity verdict, residual diagnostics (Ljung-Box, Shapiro-Wilk), forecast evaluation against the naive baseline. 5,124 daily observations of NYSE ticker A from 1999 to 2023. testthat suite, lintr config, R-CMD-check CI matrix across multiple R versions.

Rforecasttseries testthatlintr

GitHub Actions

Read the case study → View on GitHub →

2024 · Side project · Healthcare sector

Clinical Trial Data Analysis

Statistical analysis and visualisation of clinical trial outcomes in Python. Methodical walk-through: load and profile the data, handle missing values with documented rules, run hypothesis tests selected by data shape, pair every p-value with an effect size.

Python

pandas

NumPy MatplotlibStatistics

Read the case study → View on GitHub →

2024 · Side project · Aviation sector

Relational DB Design for Airport Ticketing

AirWave Express Ticketing System thumbnail with flight path arc and aircraft

SQL Server schema covering passenger management, flight scheduling, reservations, ticketing, baggage, and ancillary services. Auto-incrementing IDs via sequences and triggers, referential integrity via foreign keys, plus CHECK / NOT NULL / UNIQUE constraints. Production-grade schema design with detailed project report.

SQL ServerT-SQLSchema Design TriggersSequences

Read the case study → View on GitHub →

2025 · Side project · Engineering, statistics

Building Energy Loads

R analysis package predicting heating and cooling loads from eight building geometry parameters across 768 simulated configurations. Stepwise linear regression on each response, with VIF collinearity checks (car), Shapiro-Wilk normality, Breusch-Pagan heteroscedasticity, and the standard residual diagnostic plots. testthat suite with synthetic fixture, R-CMD-check CI.

Rstats::stepcar (VIF) testthatlintr

GitHub Actions

Read the case study → View on GitHub →

2025 · Side project · BI report, macroeconomics

Economic Resilience Dashboard

Single-screen Power BI report on 20 years of IMF World Economic Outlook data across 26 high-income economies (2001 to 2020). Star-schema model, eight indicators feeding nine DAX measures, a Metric Selector field parameter, and a Python companion validator (pandas + openpyxl + pytest) that asserts every dataset descriptor in the README against the actual xlsx contents on every push.

Power BIDAXPower Query

Python validator

pytest

GitHub Actions

Read the case study → View on GitHub →