About
I work across geoscience, software, and machine learning, building practical tools for subsurface data, interpretation and decision making.
Most of my time is spent dealing with real data that is noisy, incomplete, and occasionally uncooperative. My focus is on making that data reliable enough to support analysis, models, and people making decisions with it.
What I work on
Much of my work centres on turning messy subsurface data into something usable and trustworthy:
- Designing and maintaining workflows for well log and petrophysical data
- Using machine learning where it genuinely adds value, and leaving it out where it does not
- Building tools and interfaces that help people understand complex results
- Connecting exploratory analysis with software that can be used in production
The common thread is pragmatism. I care about data limits, method assumptions, and the context in which results will actually be used.
Experience
I have nearly 20 years of experience working with subsurface data in the oil and gas industry, across technical, software, and product-facing roles.
My work has included building and maintaining industry software modules, developing applied machine learning workflows for geoscience problems, and creating technical training material for geoscientists.
I have delivered conference workshops, written long-form technical articles, and published open notebooks focused on Python and subsurface data analysis.
That mix of hands-on technical work and software development shapes how I approach problems. I care about what works in practice, how tools are actually used, and where theory meets operational reality.
Alongside this, I write and publish independently through Subsurface Syntax, where I explore practical ways of working with geological data using Python. The focus is on clarity, data quality, visualisation, and interpretation rather than novelty for its own sake.
How I think about the work
I am cautious about hype and optimistic about tools. Machine learning is most useful when it is grounded in domain understanding and supports existing interpretation workflows rather than replacing them.
Good software design, clear visualisation, and transparent assumptions matter just as much as model metrics. A result that cannot be explained or trusted is rarely useful, no matter how clever it looks.
I am especially interested in the step between experimentation and production. How ideas move from notebooks into tools other people can rely on, and how to do that without losing rigour or context along the way.
Selected work
- Python and Petrophysics notebooks
- SPWLA Machine Learning Workshop
- Subsurface Syntax, writing on Python and geoscience