Melbourne · Tokyo
I help teams in Melbourne and Tokyo design agent systems that can do useful work inside real operations. Most of the engagement is about roles, tools, review points, and delivery discipline, so the system can survive compliance review, security questions, and day-to-day use.
Services
If you are sorting through vendor claims, internal pressure, or a stalled pilot, this is where I start. I look at the workflow, the decisions, and the handoffs so the first agent pattern is grounded in the business, not the demo.
Multi-agent flows work when each agent has a clear job, the tool access is bounded, and the escalation path is obvious. I design the orchestration, fallback paths, and review steps that keep the system usable once it leaves the lab.
The difficult part is not making an agent act once. It is making the behaviour inspectable, the data handling safe, and the approval path clear enough for a regulated team to live with.
Work
Built a search experience that could handle follow-up questions instead of stopping at keyword matches. Hybrid retrieval across OpenSearch and Azure AI Search, plus AWS Bedrock agents for deeper exploration, now runs live for art discovery.
Worked on structured clinical assessment workflows where explainability had to be part of the delivery from day one. SHAP analysis, bias review, and audit logging were included because the output needed to be reviewed, not simply generated.
Led a migration from legacy infrastructure to AWS for a major Australian bank, keeping the transition stable inside a regulated environment. That background matters because modern AI delivery depends on cloud foundations that do not wobble under pressure.
Most teams can get a proof of concept running. Projects usually stall when they move beyond demos and need to run at production scale: consistent performance under real load, clear handling when the model is wrong, reviewable outputs with proper audit logging, and enough operational trust for teams to rely on the system day to day.
Background
Working with AstraZeneca Japan on enterprise AI integration for regulated pharmaceutical and clinical workflows. The focus is practical: agentic orchestration, responsible AI controls, and delivery patterns that stand up to review.
Advisory and delivery for organisations dealing with document-heavy, regulated, or high-judgement work. Recent projects include healthcare agents, retrieval workflows, and operational automations built on AWS and Azure.
Built and operated monitoring, automation, API infrastructure, and cloud pipelines on AWS. The engineering foundation that informs current AI architecture decisions.
Led cloud transition to AWS, introduced Docker and Kubernetes, and managed delivery of major masthead sites across Nine Entertainment’s digital portfolio. Nine years of scaled web infrastructure under continuous deployment.
FAQ
Projects where the stakes are real and the system has to hold up: regulated workflow automation, multi-agent platforms, RAG search, clinical decision support, financial document processing. If it has to pass a compliance review or be explained to a non-technical executive, that is the right scope.
Because the gap is usually between strategy and implementation. Internal teams often know the business but need outside depth on model behaviour, evaluation, and governance. Large consultancies can add distance between the problem and the person doing the work.
Yes. The underlying problem is the same: building systems that can be reviewed, explained, and maintained. Tokyo work usually puts even more weight on traceability and documentation, which already sits inside my delivery approach.
Most work is scoped in phases: discovery, design, build, and handover. Some teams then keep me on as an advisor for architecture reviews, vendor decisions, or delivery checkpoints. Melbourne and Tokyo share a workable timezone window, which keeps scheduling simple.
Contact