Project 02
MSc Artificial Intelligence
A growing collection of academic notebook projects completed as part of my MSc in AI — covering fraud detection, regression, data analytics, and explainable AI. Early work, documented honestly.
Where I Am
My MSc in Artificial Intelligence is currently underway. The projects on this page are the early exercises and assignments from the programme — Jupyter notebooks built to learn the fundamentals: how to handle real datasets, apply standard ML algorithms, and critically evaluate results.
These are not production systems. They are the foundation. The jump from textbook notebook to the production pipeline in Project 1 was deliberate — I wanted to show both ends of the spectrum: what the learning looks like, and where it leads.
Notebook Projects
Binary classification on a 10k-row dataset with ~1.5% fraud rate — realistic class imbalance typical of real banking data. Applied EDA to assess feature correlations, then trained a Random Forest classifier. Evaluated with precision, recall, and confusion matrix.
View notebookA deliberate model comparison exercise: Logistic Regression vs Random Forest on the same payment dataset. The dataset turned out to be poorly representative (hence the "bad dataset" label) — a useful lesson in how data quality affects model trustworthiness.
View notebookMy first end-to-end ML pipeline — data loading, preprocessing, train/test split, model training, and evaluation. A regression problem on real estate data. Simple by current standards, but where the habit of structuring ML work properly began.
View notebookData cleaning and visualisation exercise on a global Covid-19 dataset. Handled missing continent labels caused by non-country location entries, then produced comparative continent charts using Matplotlib with logarithmic scaling to handle order-of-magnitude differences.
View notebookClassification on the Kaggle diabetes dataset (768 rows, 9 features) with a focus on explainability rather than pure accuracy. Demonstrates how to interrogate model decisions and understand which features drive predictions — a skill directly relevant to high-stakes domains like healthcare and security.
View notebookWrite-ups & Analysis
A1 Name Sorting Algorithm — Code Exploration
A2 Python vs JavaScript — Language Comparison
A3 Artificial Intelligence in 10 Years
Studies & Industry — Hand in Hand
The goal was never to finish the MSc in isolation. Each module I complete, each concept I study, I want to be putting into practice in a real engineering environment. The network anomaly detector in Project 1 is proof of that — it was built during the programme, not after it.
Landing a role in ML or AI engineering will accelerate the MSc, not compete with it. Production problems surface edge cases and constraints that academic datasets never do. That gap between textbook and real-world is exactly where the most important learning happens — and I intend to be working in it.