M.Sc. student, School of Computing, Queen's University. I work on machine learning for security: anomaly detection for industrial control systems and cloud logs, and on making ML results honest and reproducible, then shipping them as hardened, tested services.
| Project | What it is | Stack |
|---|---|---|
| ics-guardian | Unsupervised process + network anomaly detection for water-chlorination ICS (SWaT/WADI), served as a hardened cloud-native service. Fusion lifts PR-AUC from 0.39 to 0.47 on the SWaT A6 campaign. | PyTorch · FastAPI · Kubernetes/Helm · Prometheus · 6-gate DevSecOps CI |
| apple-leaf-disease-framework | Leakage-free deep-learning framework for apple-leaf disease detection with honest cross-dataset evaluation, YOLO11 detection and domain adaptation. Archived on Zenodo (DOI). | PyTorch · YOLO11 · DANN |
| cloud-threat-detection | Behavior-based threat detection on AWS CloudTrail logs with refined, human-focused alerts and an investigation dashboard. | scikit-learn · Streamlit · Docker · Trivy |
| cloud-llm-deployment-cisc886 | End-to-end LLM pipeline on AWS: PySpark on EMR, QLoRA fine-tuning of Llama 3.2 1B, Terraform infrastructure, Ollama + OpenWebUI serving. | AWS · Terraform · PySpark · QLoRA |
| infosec-rag-chatbot | Retrieval-augmented Q&A over a curated information-security knowledge base. | FAISS · Sentence-Transformers · FLAN-T5 · Gradio |
| ADDC-pytorch | Apple-leaf disease segmentation, classification and severity estimation in one pipeline. | PyTorch · U-Net · DeepLabV3+ |
- Anomaly detection for OT/ICS and cloud security
- Reproducible ML evaluation (leakage-free splits, threshold-free metrics)
- DevSecOps: CI security gates, container hardening, Kubernetes
Queen's University · GitHub