Background & approach.

Managing day-to-day analytical work to develop, refine, and maintain fraud and scam detection rules in live systems. Responding to emerging threats while balancing customer protection, customer friction, and operational workload.

Investigating fraud and scam trends across a broad range of data sources to identify gaps in existing controls and develop more effective detection strategies. Connecting analytical findings with the wider risk and business context to guide priorities.

Building classical machine learning models to strengthen rule-based detection. Translating model outputs into signals that improve how rules identify suspicious activity, supported by reproducible training, evaluation, and experiment tracking.

Sustained a fleet of five models per site, investigated production issues, and worked with subject-matter experts to keep model behaviour aligned with the physical processes being modelled.

Reviewed and improved model performance across the fleet, aligning modelling approaches with operational knowledge and strengthening confidence in production outputs.

Strengthened the engineering practices supporting model training, deployment, and monitoring. Established a roadmap for ongoing sustainment and MLOps improvements to support reliable delivery and technical continuity.

Developed a legal case-management assistant with hybrid vector and keyword retrieval, specialised employee-name search, and routing, query rewriting, retrieval, and response-generation components. Implemented an API layer, authentication, and structured logging.

Built job-design tooling supporting more than 3,000 roles. A role-profile generation workflow reduced processing time from roughly an hour to approximately 30 seconds per role, with structured Excel, Word, and PDF outputs.

Connected AI-generated outputs into existing Microsoft workflows using Power Apps, Power Automate, SharePoint, and Power BI, supporting repeatable processing and enterprise use.

Developed NLP risk-detection and agentic retrieval workflows. AI-assisted review work reduced review effort by more than 80%, while compliance automation reduced turnaround time by approximately 95%.

Contributed to responsible AI governance, workshops, mentoring, and onboarding. Worked with stakeholders to question proposed solutions and define approaches that addressed the underlying business need.

Developed analytics and models to identify equipment problems and improvement opportunities. This work identified more than $3 million in potential savings.

Worked on mineral processing empirical models, including collaboration with JKMRC at the University of Queensland, and on crusher-gap estimation. Used process data to infer physical variables that were difficult to measure continuously.

Developed ore-tonnes prediction and crusher bottleneck analytics, using Theory of Constraints to help operations understand where production was being limited.

Analysed safety data for patterns and risk factors to inform management decisions on mitigation and training.

Used data analysis, predictive modelling, and forecasting to support business decisions. Developed an understanding of data quality, interpretation, and the operational context behind analytical results.

Built business applications and automated reporting workflows, translating requirements into practical tools that reduced manual effort and supported everyday operations.

Worked with business stakeholders and technical teams to understand their needs, communicate findings, and connect analytical work with how decisions are made in practice.

Achieved multiple months of 100% customer satisfaction while handling complex and sensitive customer issues.

Accredited to provide general financial-product advice. Built a working understanding of banking products, customer needs, risk, and compliance boundaries.