Projects

edgewords is a daily social word puzzle game that blends Wordle-style brevity with Scrabble-like wordcraft, built around geometric puzzles that require chaining words across shape sides. Players solve four puzzles per day (triangle to hexagon), chase par solutions, and compete via leaderboards, groups, friends, and achievements. The platform includes native iOS and Android apps with cross-platform sync, a FastAPI backend with SQL storage, and a growing social layer for messaging, streaks, and community play.

The Story Chamber is a literary innovation platform that showcases emotionally resonant, narratively coherent fiction generated entirely by artificial intelligence. Unlike AI writing tools that assist human authors, this platform constructs deeply detailed AI author personas, each with a unique backstory, worldview, and voice, that produce unedited fiction guided by sophisticated narrative blueprints. The result is machine-generated literature that feels authored, not assembled. With an emphasis on genre diversity, artistic integrity, and scalable storytelling, The Story Chamber reimagines what authorship means in the age of generative media. It offers readers a new kind of literary experience: fiction crafted by simulated minds, designed with care, and delivered with intent.

StomaGAN represents a sophisticated deep learning framework that leverages Generative Adversarial Networks to address the critical challenge of limited training data for fine-grained visual analysis. This research demonstrates how a modified DCGAN architecture, enhanced with adaptive top-k selection, spectral normalisation, and strategic noisy labels, can generate photorealistic synthetic imagery from minimal datasets. The approach reduces training instability while producing high-fidelity data that improves downstream model performance. The synthetic data has been validated through training deep CNNs for automated detection, achieving strong accuracy on unseen real-world imagery and establishing a reusable blueprint for data augmentation in computer vision.

StomataHub is a collaborative ecosystem designed to democratise access to AI datasets, reproducible pipelines, and applied machine learning tools. The platform serves as a centralised repository for annotated datasets, peer-reviewed research, and production-grade ML applications. It features a web-based detection system under active development, incorporating modern computer vision architectures and scalable inference workflows. By fostering interdisciplinary collaboration and open science, StomataHub accelerates research and model development, while providing a practical template for transparent data management and applied analytics.

This research introduces a deep learning methodology for high-accuracy detection and morphometric analysis in high-resolution microscopy imagery. The project presents a custom convolutional architecture with advanced feature extraction, achieving accuracy above 98% across diverse imaging modalities. The framework enables high-throughput quantitative analysis of density, morphology, and spatial patterning, supported by a carefully annotated dataset of thousands of images. The work bridges domain science and applied AI, demonstrating robust, transferable computer vision techniques for complex visual data.

This research program advances autonomous vision systems through the integration of computer vision, machine learning, and optimized path planning. The work introduces multi-objective algorithms that maximize information acquisition while minimizing acquisition cost, paired with hybrid volumetric and point cloud reconstruction for high-accuracy 3D modeling. GPU-accelerated pipelines enable near real-time performance, providing a practical foundation for scalable 3D perception systems and robust spatial analytics.