Time-series forecasting • Optimisation foundations (LP/MILP concepts) • Python/R (pandas, scikit-learn, PyTorch) • SQL • BI (Power
BI/Tableau) • Cloud (AWS; quick to adopt Databricks/Spark) 5+ years turning messy, multi-…
Time-series forecasting • Optimisation foundations (LP/MILP concepts) • Python/R (pandas, scikit-learn, PyTorch) • SQL • BI (Power
BI/Tableau) • Cloud (AWS; quick to adopt Databricks/Spark) 5+ years turning messy, multi-source data into production-ready models,
tools, and dashboards that improve on-time performance–style KPIs, reduce manual reporting ~30%, and translate insights into clear
actions for operations and commercial leaders.
Lead data scientist on a Queensland Cancer Registry project: built reproducible Python/R pipelines for cohort assembly, data validation, and feature engineering across large linked datasets.Delivered survival/risk-stratification models to support treatment planning; documented assumptions, calibration, and fairness checks for decision-makers.Shipped executive-ready dashboards and briefs (Power BI/R Markdown) that turned model outputs into clear pathway recommendations and measurable actions.Established governance artifacts (data lineage, ethics/consent mapping, code reviews) to keep analytics compliant, auditable, and handover-ready.Coordinated clinicians, statisticians, and consumer reps; ran short delivery cycles with defined milestones and stakeholder check-ins.
Lead data scientist on a Queensland Cancer Registry project.
Built reproducible Python/R pipelines for cohort assembly, data validation, and feature engineering across large linked datasets.
Delivered survival/risk-stratification models to support treatment planning.
Documented assumptions, calibration, and fairness checks for decision-makers.
Shipped executive-ready dashboards and briefs (Power BI/R Markdown) that turned model outputs into clear pathway recommendations and measurable actions.
Established governance artifacts (data lineage, ethics/consent mapping, code reviews) to keep analytics compliant, auditable, and handover-ready.
Coordinated clinicians, statisticians, and consumer reps.
Ran short delivery cycles with defined milestones and stakeholder check-ins.
Built end-to-end pipelines for multi-source ICU data (EHR, bedside monitors, labs): ingestion → validation → feature store; improved downstream modelling speed and reliability.Developed and validated ML/DL models for ICU outcomes (mortality/deterioration), HRV/ECG ectopic-beat detection, and readmission risk; focused on calibration, interpretability, and harm minimization.Containerized training/inference (Docker) and automated retraining/reporting on AWS (S3, Athena, SageMaker); added logging/metrics for traceability and cost awareness.Produced decision dashboards and “so-what” narratives for clinicians and operations, reducing manual reporting effort by ~30% and increasing trust in KPIs.Implemented privacy-preserving workflows for clinical text (de-identification) and robust QA checks (schema tests, missingness/outlier monitoring).Collaborated daily with engineers, clinicians, and product leaders; turned vague questions into measurable use-cases, features, and success criteria.
Assisted in the development of AI and machine learning models for ICU and oncology projects. Annotating unstructured pathology reports and building hierarchical knowledge-base ontologies using machine learning algorithms.Conducted data preprocessing, feature engineering, and statistical analysis on large-scale clinical datasets, ensuring high data quality and reproducibility.Supported grant and project proposal preparation by conducting literature review, research synthesis and technical documentation.Work with multidisciplinary teams of clinicians, data scientists, and industry partners, bridging the gap between technical innovation and clinical application.Delivered presentations and research dissemination at academic meetings and conferences and workshops, strengthening institutional visibility and collaboration opportunities.Trained and mentored junior students in data science, coding, and research methods, fostering skill development within the research group.
Built end-to-end pipelines for multi-source ICU data (EHR, bedside monitors, labs): ingestion → validation → feature store.
Improved downstream modelling speed and reliability.
Developed and validated ML/DL models for ICU outcomes (mortality/deterioration), HRV/ECG ectopic-beat detection, and readmission risk.
Focused on calibration, interpretability, and harm minimisation.
Containerised training/inference (Docker).
Automated retraining/reporting on AWS (S3, Athena, SageMaker).
Added logging/metrics for traceability and cost awareness.
Produced decision dashboards and “so-what” narratives for clinicians and operations.
Reduced manual reporting effort by ~30%.
Increased trust in KPIs.
Implemented privacy-preserving workflows for clinical text (de-identification).
Implemented robust QA checks (schema tests, missingness/outlier monitoring).
Collaborated daily with engineers, clinicians, and product leaders.
Turned vague questions into measurable use-cases, features, and success criteria.
Developed advanced queries using SQL to set up and configure an ETL framework to manage data on the data warehouse.
Improved report loading time by 15%.
Optimised server performance.
Conducted programmatic validation and cleansing for data listing reported on clinical trial data.
Introduced process and system improvements.
Led to a 25% increase in accuracy and consistency of data within 3 months.
Implemented and maintained SAS programs.
Implemented and maintained safety analysis datasets, safety tables, and listings used for analysis and reporting of clinical trial data.
This project aims to fill key gaps in current colorectal cancer (CRC) care by integrating advanced analytics with patient-centred design. First, we will apply innovative statistical methods to large-scale cancer registry data to understand how comorbidities influence clinical outcomes. Second, we will develop advanced AI models to improve prognostic and predictive risk-stratification capabilities for CRC patients with multiple health conditions. Finally, we will engage patients, clinicians, and stakeholders to co-design the first clinical tool that extracts and transforms comorbidity and clinical information into an MDT-ready format, supporting more informed, personalised, and efficient treatment decision-making.
Developed an end-to-end framework for dynamic risk modelling in TBI during the initial ICU stay using deep learning techniques. First, mapped physiologic–clinical interdependencies using correlation-based network analysis coupled with graph neural networks to surface key variables and communities. Built a deep neural model for ectopic-beat detection from ECG and assessed generalisability with both internal and external test sets. Designed a hybrid prognostic model that combines a custom learned-weight module with a BiLSTM unit to capture temporal dynamics. Finally, delivered a continuously updating ANN that ingests hourly physiological time-series to generate real-time trajectories of 30-day mortality risk. Together, these components provide interpretable drivers and streaming risk estimates to support ICU decision-making.
This project aimed to develop an explainable, AI-driven clinical decision-support system for personalised cancer treatment planning. Many cancers, including breast cancer (BC) and colorectal cancer (CRC), demonstrated heterogeneous treatment responses due to complex patient- and therapy-related factors, meaning that treatment effects varied widely across individuals. Although patient-centred cancer care had advanced, clinical guidelines still largely prioritised tumour characteristics and often overlooked broader contextual factors. Existing AI tools offered promise but predominantly relied on opaque “black-box” models and were mainly focused on diagnosis, limiting their clinical usefulness because of ethical, interpretability, and trust concerns. In this project, we built a transparent “open-box” AI system that learned from historical patient data and provided clinicians with clear, evidence-informed treatment recommendations. The system integrated a knowledge base, computational optimisation algorithms, and multimodal patient information—including cancer stage, pathology reports, comorbidities, and socio-demographic factors—to support personalised decision-making. Clinician expertise was embedded through co-design, ensuring that the system aligned with real-world decision pathways and continuously improved in performance and clinical relevance.
نظام التوظيف التشغيلي المعتمد على الذكاء الاصطناعي أولاً، الذي يوحّد ويؤتمت دورة التوظيف بالكامل للمؤسسات الحديثة حول العالم.
د.جوب منطقة حرة ذ.م.م. 2026 © جميع الحقوق محفوظة
