Showcasing a selection of data science and engineering projects.

Bridge Condition Regression Project

EDA | Linear Regression | Data Visualisation

This detailed project investigates the condition of bridges in Texas using a dataset provided by the US Department of Transport. The goal is to analyse the predictors ability to predict the bridges condition and develop a linear model.

Audio Deception Detection Project

Machine Learning | Data Mining | Classification

This project aimed to develop an ensemble machine learning model to predict whether a human read audio story is true or false. This supervised binary classification task was applied to the MLend Deception data set - to which I was a contributor.

CIFAR10 Classification with Novel Architecture

Neural Network Training | Architecture Design | Classification

In this project I built a fully working academic search engine combining classical information retrieval with modern BERT language models. Enabling true semantic search across interdisciplinary scientific literature. It was evaluated against a range of variants and with BM25 as a baseline - using cutting edge 'AI as a judge' for relevance judgements.

This attempt at the popular CIFAR10 dataset acheived 92.06% accuracy, approaching state-of-the-art performance without using transformers or attention mechanisms. Instead I implement adaptive pathway weighting through SoftMax normalization, creating a kind of attention mechanism that focusses on the most relevant features through pathway weightings .

Latent Semamtic Search Engine

Information Retrieval | Finetuning | BERT

Portfolio

AI Assisted Minesweeper Game

Web Development | Azure | NextJs | Docker

This hobby project is a Azure deployed version of the classic Minesweeper game - but with a twist. If you get stuck, the AI assistant can analyse the board and make the statistically safest move for you.