MSc Artificial Intelligence graduate and Machine Learning Developer working across medical image analysis, deep learning, reproducible experimentation, and responsible model evaluation — with a growing focus on oncology-oriented clinical AI.
My work is directed toward medical AI for oncology, especially where imaging, clinical data, and responsible model evaluation can support better diagnosis, risk assessment, treatment planning, and follow-up decisions.
My MSc dissertation on medical image concept detection gave me a foundation in radiology image pipelines, multi-label deep learning, and clinically meaningful evaluation beyond accuracy. I am building on this toward oncology imaging, breast cancer AI, radiomics/radiogenomics, and reproducible clinical AI workflows.
Selected work and ongoing projects in medical imaging AI.
Built a ResNet18 multi-label classifier on radiology images with UMLS/CUI concept labels using ImageCLEF/ROCO-style data. Evaluated using per-concept precision, recall, and F1-score with detailed analysis of label noise and evaluation strategy.
A structured machine learning pipeline for breast cancer classification using the Wisconsin Diagnostic Breast Cancer dataset. Covers preprocessing, baseline model comparison, and evaluation beyond accuracy — with documented model limitations and a roadmap toward imaging-based breast cancer AI.
A clean medical image classification pipeline using MONAI and PyTorch on the MedNIST dataset. Focused on reproducible preprocessing, validation metrics, confusion matrix analysis, precision, recall, and F1-score. Serves as a reproducible template for future oncology imaging workflows.
A research-grade medical imaging AI project developed toward oncology-oriented concept detection and clinically meaningful model evaluation. The current phase establishes a reproducible chest X-ray classification baseline using PyTorch, with structured dataset handling, model training, test-set evaluation, confusion-matrix analysis, and prediction visualisation. The first baseline uses pneumonia-vs-normal chest X-ray classification to validate the imaging workflow and examine clinically relevant trade-offs beyond aggregate accuracy. Current results show high sensitivity to pneumonia cases, while also highlighting the cost of over-flagging normal cases — making the project a practical starting point for studying false negatives, false positives, threshold behaviour, and clinical risk in medical AI systems. The project will next progress toward oncology-relevant imaging tasks, with planned extensions in lung imaging, radiomics features, explainability, and clinically framed model validation.
The themes that connect my current work and research direction
Deep learning and quantitative workflows for radiology and cancer imaging data.
Screening, imaging-based risk assessment, and responsible evaluation of AI tools in clinical workflows.
Quantitative imaging biomarkers and links between imaging, clinical, and biological data.
AI systems that support diagnosis, risk assessment, treatment planning, and patient follow-up.
Interpretable and responsible model evaluation frameworks for healthcare contexts.
Careful metrics, validation workflows, and documentation for clinically meaningful assessment.
Investigating Concept Detection Techniques in Medical Images — Heriot-Watt University, 2023.
View profile ↗Research-oriented computational validation across complex software systems: translating platform-specific behaviour into testable implementation logic, comparing outputs, debugging discrepancies, and documenting reproducible workflows under confidential project constraints.
University–industry ML research internship. Implemented and compared optimisation algorithms including genetic algorithms, Bayesian optimisation, PSO, SMAC3, and Dragonfly, with international collaboration, research documentation, and supervisor-led progress reviews.
Supported undergraduate Java programming labs and postgraduate ML sessions. Mentored students through debugging, assignment queries, and model workflow concepts.
Managed sensitive customer and financial documentation in a regulated, audited environment. Developed skills in traceability, compliance-aware execution, and structured quality control — transferable to clinical AI governance and responsible deployment in regulated healthcare settings.
Java enterprise development and automation testing for an e-commerce client. Supported issue investigation, Agile/Scrum delivery, and data migration from MySQL to Redis via Cassandra.
Dissertation: Investigating Concept Detection Techniques in Medical Images. Modules: Machine Learning, Research Methods, Biologically Inspired Computation, Big Data Management, Distributed and Parallel Technologies, Advanced Software Engineering.
Relevant modules: Artificial Intelligence, Machine Learning, Data Mining, Algorithms, Databases, Software Engineering.
Particularly interested in groups working on breast cancer AI, cancer imaging, radiomics/radiogenomics, clinical decision support, and reproducible evaluation of AI systems in healthcare — worldwide.