Chartered Accountant, Data Insights
Hello and welcome to my website! Here you will find projects I have done to solve problems applying business acumen, financial discipline & rigor, and data science techniques.
Every good story has a quest. Mine is to make sure that people’s time, and the data they collect, are spent on work that matters. It isn’t finished yet, so here is how it has gone so far.
It began the way many stories do, with a graduation. I walked out with a bachelor’s degree and into the world of finance, keen to be useful. And I was. Over the years I solved one problem after another, and I loved being the person who made things work.
One problem kept following me, though: highly trained accountants spending their valuable time on menial work. The idea struck while I was helping implement a new finance system at the New Zealand Law Society. If I wanted to fix this, I had to understand an accountant’s work from the inside, then go looking for tools that could take the tedium away.
So I left the Law Society and set out to earn my chartered accountant qualification. It was a leap into the unknown, and it gave me the skills, and the credibility, to see the work the way accountants do.
Newly qualified, I helped on a financial management information system (FMIS) upgrade, developing the project financial module. I was convinced that tools to automate the no-value-added work already existed. What was missing was a bridge between the people building the tools and the business users who needed them. I decided to become that bridge, and trained as a project manager. My journey took a new path.
Then I began uplifting the Power BI capability of a finance analysis and reporting team, and I saw something bigger. We collect so much data nowadays, and yet we still feel we don’t have enough information to make decisions, so gut feel fills the gap. We have an arsenal of tools for producing reports, and little time to study what the reports are telling us. So I left New Zealand Police and went back to study data science.
I have completed that study, and I’m carrying new tools: statistical modelling and data science disciplines, alongside accounting rigour and business know-how. I’m ready to take on the data problems we face today. The quest continues, and the projects on this site are the first steps.
14 July 2026
Problem: Can we predict how long an IT incident will take to resolve, to plan staff and set realistic expectations?
Approach: Reshaped 141,712 events into 24,918 incidents, then compared linear, multiple and lasso regression models.
Result: The multiple regression model explained 89% of the variance in resolution time, with predictions off by 7–8 days on average. Each system change added about 2 days.
27 July 2026
Problem: Can we predict machine failures from sensor readings, to focus maintenance where it matters?
Approach: Explored 944 sensor observations, compared six classification algorithms, then built a random forest model.
Result: The random forest predicted failures with 92.6% accuracy. Volatile organic compounds (VOC) and air quality were the strongest predictors.
14 August 2026
Problem: Can machine learning forecast near-term spending early enough to warn finance teams of budget overruns, instead of manual Excel forecasts?
Approach: Used a government invoice register as a proxy, and compared classical time series models with LSTM deep learning models on contractor and labour hire costs.
Result: SARIMA forecast contractor costs best and a two-layer LSTM forecast labour hire costs best. Forecasting is a viable early-warning signal for sustained trend shifts.
I’d love to hear from you. Whether you have a data problem to dig into, a project to talk through, or just want to say hello, the best way to reach me is on LinkedIn.
A case study on machine learning · Presented 14 July 2026
Tip: use the left and right arrow keys to move between slides.
A case study on predicting machine failures · Presented 27 July 2026
Tip: use the left and right arrow keys to move between slides.
A case study on predicting business spending · Presented 14 August 2026
Tip: use the left and right arrow keys to move between slides.