Research
Clinical questions, statistical methods, and where the work landed.
Screening data combine different modalities, referral patterns, clinical contexts, and population-level inequities.
Risk modelling, population cohort analysis, explainable imaging AI, and deployment-aware validation.
Led the analysis for the 30,000 Voices national breast cancer report, which shaped ~half of NZ's national breast cancer KPIs.
Embryo assessment is information-rich, time-sensitive, and difficult to make consistently measurable.
Computer vision turns time-lapse imagery into traceable morphology, timing, expansion, and transition measurements.
Patent-pending WyldBloom decision-support tools, international validation, and clinic-pilot workflows.
Embryo selection is subjective, time-pressured, and burdened by manual paperwork that scales poorly across clinics.
Patent-pending cohort-relative embryo ranking on a 45K-embryo, hardware-agnostic dataset, plus lab automation reducing transfers and administrative load.
Patent-pending embryo ranking and lab automation tools now running in NZ fertility clinics, reducing transfers and reclaiming embryologist time.
Biomedical results lose value when the path from data to conclusion cannot be inspected or repeated.
Reproducible workflows, transparent validation, interactive statistics tools, and explicit assumptions.
Open-source practice and teaching resources for PCA, collinearity, regression, and applied research.