In this episode of Whimsical Wavelengths, we explore how machine learning is opening new possibilities in geophysical imaging and inversion workflows. Like image segmentation! We look at how modern computational tools can help interpret what we cannot observe directly beneath the surface.

Our guest, Johnathan Kutti, joins us to break down how machine learning approaches can assist with geophysical inversion, improve subsurface models, and support decision-making in exploration and environmental studies. With experience both in the field and in building mathematical tools, he brings a grounded perspective on how these methods work in practice.

We start by outlining what geophysics actually is—using physics to study the Earth’s structure and processes—and why inversion methods are so central to the field. Because we cannot directly measure physical properties everywhere inside the Earth, geophysical inversion works backward from measurable data such as magnetics, gravity, or electromagnetic responses to estimate what the subsurface must look like.

The conversation then moves into:

• Why geophysical inversions have infinite possible solutions
• How physical assumptions and constraints narrow those solutions
• Where machine learning and image segmentation can help
• Examples of integrating AI into geoscience workflows
• Practical realities from years spent collecting data across remote terrain

If you've ever wondered how AI and scientific modeling intersect—or how we “illuminate the void” geophysically—this episode offers both clarity and depth.

UBC Geophysical Inversion Facility: https://gif.eos.ubc.ca/

Whimsical Wavelengths Links

Facebook:www.facebook.com/WhimsicalWavelengths

Instagram: @whimsical.wavelengths

Bluesky: @whimsicallambda.bsky.social

Email: whimsical.wavelengths@gmail.com

Patreon: patreon.com/WhimsicalWavelengths