Data Science Projects

Academic work only counts for something if the model is honest about its limits instead of leaning on its headline accuracy. Each of these projects had a number that looked good, and a reason that number could not be trusted on its own.

Four projects taken past the notebook stage. They sit together here because they are one body of academic work rather than four separate products: a Twitter/X bot-detection classifier, a customer churn model, an R Shiny dashboard on graduate employment, and a hackathon submission on sovereign debt.

Python with Pandas and scikit-learn for the modelling, R and R Shiny for the interactive dashboard. The work covered cleaning real data, engineering features, training and evaluating classifiers, and reporting what the results did and did not support.

• Twitter/X bot-detection classifier
• Customer churn prediction model
• R Shiny graduate-employment dashboard
• 10Alytics Global Data Hackathon sovereign-debt analysis, among the 217 reviewed from 2,282 participants across 51 countries

For the 2025 10Alytics Global Data Hackathon, I completed a full sovereign-debt analysis alongside 2,282 participants across 51 countries, and was among the 217 reviewed. I kept the write-up to the conclusions the data actually carried.

A completed set of academic projects from the Eduvos BSc. They are learning projects, and they are framed as such. They are also where I picked up the habit of checking what the data really supports, which is the part that carried into production work.