Mapping the Unseen: How Sam Thiele brings Machine Learning into Geoscience
Sam Thiele
In the second season of Keeping up with the Boost Fellows, GSO’s Birte Seffert travels across Germany to visit the 15 fellows who started their Klaus Tschira Boost Fund projects in 2025. The series offers a closer look at their research, their working environments, and the realities of building an academic career today – across disciplines, institutions, and career stages.
The first stop: Freiberg, Saxony. Here, Birte meets Sam Thiele, project and team lead at the Helmholtz Institute for Resource Technology, working at the intersection of geoscience, machine learning, and data integration — an example of how machine learning in geoscience is increasingly shaping the field.
After completing his PhD in Australia in 2019, he moved into academia in Germany. This move was the result of a conference conversation that led to a position in Freiberg, a village he had never been before but felt familiar with having grown up in rural Australia.
What does it take to understand what lies beneath our feet, and why does it matter?
Resources like metals, energy, and water all come from underground. And whether we can mine or manage these resources responsibly depends on how well we understand what is actually where. This is especially critical given the ongoing energy transition: one wind turbine can require the mining of as much as 3,000 tons of rock. The implications for landscapes and communities due to mining are huge, so finding resources in appropriate places and extracting them efficiently is more important than ever.
Sam´s Boost-funded project focuses on improving the three-dimensional geological models that predict the distribution of rock types underground. These models are essential for both identifying, managing and efficiently using these underground resources, and understanding associated risks or hazards, but are also inherently uncertain, Sam says, since we only have limited data. Interpretations from different geologists using the same data can differ by a surprising amount.
His approach: combining hyperspectral imaging with neural networks that objectively integrate diverse data to reduce this uncertainty. The goal is not just more precise models, but more informed decisions – about resources, environmental impact, and risk.
Between fieldwork, algorithms, and reality checks
Walking through the institute with Sam quickly reveals how complex this work is in practice. Hyperspectral cameras worth millions of euros and classified as dual-use technology come with strict regulations. Drones used for fieldwork require logistical workarounds – the batteries cannot simply be shipped due to the risk of explosion.
Sam’s role sits right at the interface between fundamental research, technological development, and real-world constraints like paperwork, permitting and the distance a mule can carry kilos of equipment across rough terrain.
Interdisciplinarity: beyond the buzzword
Sam stresses that geoscience is inherently interdisciplinary. Physics, chemistry, biology, and computer science all come together when trying to understand the Earth.
But working across disciplines is not just about combining methods – it is about communication.
“There’s no way around spending quite a lot of time building a common language. Otherwise, you simply can’t talk about the details properly.”
He also points to a tension many researchers will recognize:
“There’s the marketing of interdisciplinarity – and then there’s actually doing it.”
Building networks, investing time in shared understanding, and finding the right collaborators are essential – but often underestimated parts of the work.
A breakthrough and the role of risk
Sam experienced what cannot be planned: a breakthrough.
Together with a team member funded through the Boost project, he developed a new computational representation for geological structures – more flexible than existing approaches and opening up multiple future research directions.
“The ideas were high-risk. There was a large chance it wouldn’t work.”
This is where the Boost Fund plays a specific role.
“It’s sort of a seed funding. It allows you to try something new. And if it doesn’t work – that’s also okay. You are allowed to fail.”
TOP 3 career advice from Sam Thiele
- Be aware of the gap between “interdisciplinary” as a label and as practice
“It’s much easier to pretend to be interdisciplinary than to actually do it.”
Real interdisciplinary work requires depth in at least one field, plus the effort to translate, integrate, and follow through — not just positioning.
- Build your network early – and invest in it
Interdisciplinary work depends on people who understand each other. Building that takes time.
- Don’t be afraid to try things that might fail
“Science is about being wrong in a useful way. Scientists should fail and fail fast and fail often. You may need to try lots of different things until you find something that works. The challenge is thus to construct failures in a way that you learn something useful. ”
Why the Klaus Tschira Boost Fund matters
Sam describes the Boost Fund as filling a gap in the funding landscape – especially for researchers who are not yet eligible for large-scale grants or who fall outside standard funding schemes.
“It gives you the flexibility to explore ideas and develop your own direction. That’s something that’s often missing.”
About the KT Boost Fund
The Boost Fund supports postdoctoral researchers and early group leaders in Germany with flexible funding for independent, often higher-risk and interdisciplinary projects, combined with career development opportunities and access to a strong peer network.
The program addresses a critical phase after the PhD, where researchers are expected to develop independence, but often lack the resources and flexibility to do so.