Selected work / Risk · Analytics · Applied AI
A portfolio of practiceThe stories behind the systems.
Real problems, things I built, and what changed. A collection of work turning complexity into something people can act on.
Explore the collectionEngineering instinct.
Decision-making impact.
Built in practice. Used by people. Designed to last.
Butterfly
I proposed and built Butterfly after spotting delays and errors in manual risk reporting. I traced the source data and translated written risk requirements into measurable thresholds.
Connected source-based risk monitoring with measurable policy thresholds and Board reporting.
Rapier
Working alongside auditors, I turned testing procedures into SQL rules that could examine the entire journal population.
Replaced manual sampling with 12M+ automated journal tests.
AI-Assisted Reporting
I built and rolled out a ChatGPT-assisted workflow combining risk data and external news into branch report drafts. Risk owners review and finalise the narrative.
More branches supported, less time preparing reports.
Automation Discovery & Prioritization
I led two colleagues in reviewing approximately 120 insurance processes, presented opportunities to management, built a selected pilot and transferred knowledge to the internal team. Explore the approach through two illustrative industry assessments.
Built an opportunity library and a selected pilot, then transferred knowledge to the internal team.
Selected controls from audit practice, reconstructed with sample evidence.
Controls in Practice
Duplicate invoices. Unsupported receipts. Aging stock. Overlapping duties. Four controls I selected to show how business context and follow-up evidence shape an audit conclusion.
Scenario work for decisions that can’t wait for certainty.
Full-population lead-time and inventory ranking — which suppliers and regions drove delay after the flood.
Time to Survive vs Time to Recover. Move the disruption window to explore the trade-off.
Explore the full scenarioConnect external news to potential partner risks, then review the evidence and risk score with the owner.
How much could shared components save, and what happens when the supplier stops delivering? Compare three options and prepare a management discussion.
Explore the thinking and try the models. All interactive demos use illustrative data.
Field notes from the work between the big projects.
A team was waking before dawn to prepare reports before internet access disappeared. I built a small automation around the tools they already had.
Years later, a former colleague told me it was still running.
The team had a problem that had become normal.
Because of the situation in Myanmar, internet access could be cut off during the day. The team needed to wake up very early, run queries, prepare reports, and get everything done before they lost connectivity.
I spent time understanding how the team actually prepared the reports — what they ran, what data was available, and what tools we already had.
At the time, Google Apps Script was one of the practical tools available to us, so I built a small automation around the existing process.
The reports could be prepared automatically and be ready when the team woke up.
No new platform. No complicated architecture. Just a small piece of automation solving a very real problem.
Years later, I was surprised to hear that the solution was still there and still working.
Sometimes good engineering is simply paying enough attention to notice what shouldn't be happening in the first place.
The thread through it all
Understand the problem.
Build something that helps.