Making Computers Smarter to Understand Boiling, Blubby Flows – Dr. Amra Hasecic on Machine Learning, Coffee, And Why Understanding Flows Has Universial Benefits For Our Planet & Society
“I’m working on making computers smarter so they can rapidly comprehend and predict the behavior of hot, bubbly flows – essentially creating a turbo-charged shortcut for scientific discovery,” Dr. Amra Hasecic explains.
Welcome to our latest episode of the In Her Element interview series, where we sit down with inspiring WiRe Fellows to talk about their research, routine, and what it means to be a women in research today. In this conversation, WiRe Fellow Dr. Amra Hasecic shares her passion for unraveling the mysteries of complex flows and their far-reaching implications. By teaching AI to improve our understanding of these intricate phenomena, she aims to optimize various production and energy processes, ultimately contributing to a more sustainable future. Let’s dive in and learn more about what “Double C” and bathtubs have to do with Amra’s cutting-edge research.
Dr. Amra Hasecic
Researcher in Machine Learning in Fluid Dynamics
Host Institute: Mathematics Münster (Cluster of Excellence)
Scientific Host: Prof. Dr. Mario Ohlberger
Research Project
Machine learning for multiphase flows with radiative heat transfer – MANIFOLD
Amra’s research helps …
…to optimize various industrial processes, decrease energy usage, and contribute to a more sustainable future for all.
What and why: From a Desire to Achieve Social Justice to Gender, Cultural, and Critical Race Studies
What is the your core research question – in one sentence?
Can we use machine learning to truly improve computational fluid dynamics in the field of multiphase flows with radiative heat transfer or the current impact is just not high enough.
Ok:-). And now for all: How would you explain your research to a child?
Imagine you’re playing with a bathtub full of water.
If you throw in bubbles, or oil, or even sand, the water doesn’t look simple anymore—it has many things mixed together. That’s what a multiphase flow is: water plus bubbles, or other stuff moving around together.
Now, imagine shining a flashlight into the bath. The light gets bent, bounced, or blocked by the bubbles and drops. That’s like radiative heat transfer—heat moving around in the form of light and radiation, not just by touch.
The problem is: when scientists try to calculate exactly how all this water, bubbles, and light move together, it’s like trying to count every bubble and every sparkle of light in the tub—way too slow for even the best computers.
So what I do is teach a computer (with machine learning) to “guess” the answers much faster, by learning from examples. It’s like teaching the computer to recognize patterns like ‘when bubbles are this big, the water moves like that’, or ‘when the light shines here, it heats up there’
This way, instead of waiting days for the computer to finish, we can get answers in hours or minutes—and still be very close to the truth.
In short: I make computers smarter so they can quickly understand and predict how hot, bubbly flows behave—like a super-fast shortcut for science.
Can you share what first sparked your interest in this field?
From the very beginning of my scientific career, I have been focused on pushing the boundaries of Computational Fluid Dynamics (CFD). One of my first achievements was developing a mathematical model and numerical method capable of accurately solving multiphase flow problems coupled with radiative heat transfer and phase changes. Up to that point, radiative heat transfer in such problems was either neglected or treated in an extremely simplified manner. Considering that radiative heat transfer is the dominant heat transfer mechanism at high temperatures, this was far from a negligible simplification. However, one of the major drawbacks of CFD has always been its extremely high computational demands and the limitations that arise from them.
With the rapid expansion of machine learning and computational capabilities over the past few years, I naturally began to wonder whether we could leverage these advances to address what I already knew to be a highly computationally demanding problem.
Can you describe a key moment or turning point in your research?
A key moment occurred when I have realized that due to the highly nonlinear and implicit nature of the equations, conventional approaches are not sufficient and one need to think outside of the box for this one 😊.
What would you love to understand or solve within the next five years?
Over the next five years, I would love to develop machine learning methods that can reliably accelerate simulations of multiphase flows with radiative heat transfer. My goal is not only to reduce computational costs, but also to deepen our understanding of how data-driven models can capture highly nonlinear and coupled physical phenomena without losing accuracy. Ultimately, I want to create tools that make these complex simulations accessible for broader use in science and industry—helping engineers and researchers solve real-world high-temperature flow problems much faster and more efficiently.
Inside the Research Routine: A Day in the Life of a Researcher in Machine Learning
What does a typical research day look like for you?
I usually begin by running or checking simulations, making sure the data looks consistent, and identifying where adjustments are needed. The middle part of the day is often spent testing machine learning models, comparing results, and troubleshooting why something does—or doesn’t—work. There’s also a lot of reading involved: keeping up with the latest papers: in the field of data-driven fluid mechanics, new achievments are on a daily basis — so you really need to keep up.
But research isn’t only about the computer. I also spend time discussing ideas with colleagues, mentoring students, searching for suitable funding calls to support my research in future. And, of course, every day brings a bit of the unexpected—an error message, a breakthrough, or just a new question that leads me down another path.
In short, my days are a mix of coding, thinking, learning, connecting and sharing.
Do you have any morning rituals that help you get into the research mindset?
Double C: Coffee and Code.
Where and when do you have your best ideas?
One of the best ideas came to me when I was not actively thinking about the problem. I have a feeling that my mind is always working, like a background task on a computer. And very often, when I meet certain challenge and I spent a lot of time brainstorming about the solution, just stepping out of that mode and spending time in different block of a day would lead to an “aha” moment and possible solutions.
How do you recharge when you hit a research block?
My day is made up of several overlapping blocks: research, teaching, and spending time with my kids. Whenever I feel oversaturated in one of these areas, it usually coincides with the natural time to switch to another. So I don’t really have a “usual recipe” for recharging. If I had to put it simply, I would say that my family is my greatest source of energy and recharge.
Your Field – Then, Now & What’s Next: Navier-Stokes Equations, Flows, And Why Amra Would Love to Have Dinner With Ludwig Prandt
Which historical figure in your field would you like to have dinner with – and what would you ask?
One historical figure that comes to mind is Ludwig Prandtl. He’s considered the father of modern fluid mechanics because of his boundary layer theory, which completely reshaped how we understand fluid flows.
At dinner, a fascinating question to ask him would be: If you had access to today’s computational power and machine learning tools, how would you rethink or extend your theories?
Ludwig Prandtl, a German engineer and physicist, introduced the boundary layer theory in the early 20th century. He described the behavior of fluids near solid surfaces, where the fluid’s velocity increases from zero to its free-stream value. This thin region, the boundary layer, affects drag, lift, and heat transfer.
Prandtl’s theory divided the boundary layer into laminar and turbulent regions, enabling engineers to design more efficient shapes for aircraft, ships, and vehicles. His work had a profound impact on aerodynamics, hydrodynamics, and chemical engineering, and remains a fundamental concept in fluid mechanics today.
What do you consider the greatest achievement in the history of your discipline?
I would consider the development of the Navier–Stokes equations as the greatest achievement.
Navier-Stokes Equations
These equations are named after Claude-Louis Navier and George Gabriel Stokes, who first formulated them in the 19th century.
What do they describe?
The Navier-Stokes equations describe how fluids (like water, air, or gases) move and interact with their surroundings. They help us understand and predict the behavior of fluids under various conditions, such as pressure, temperature, and velocity.
In simple words:
Let’s go back to Amra’s bathtub example. Imagine you’re throwing sand into a bathtub full of water. The ripples that form and spread out when you throw the sand in are like the fluid flow described by these equations. They take into account factors like:
How fast the fluid is moving (velocity)
How thick or viscous the fluid is (sand-water mixture vs. water)
The forces acting on the fluid, like gravity or pressure
The boundaries of the fluid, like the edge of the bathtub
Mathematically speaking…
The Navier-Stokes equations are a set of nonlinear partial differential equations that relate these factors to each other. They consist of four equations:
1. The continuity equation (conservation of mass)
2. The momentum equation (Newton’s second law applied to fluids, x-direction)
3. The momentum equation (Newton’s second law applied to fluids, y-direction)
4. The momentum equation (Newton’s second law applied to fluids, z-direction)
What impact has this discovery had on society – or on you personally?
They represent a universal framework for describing how fluids move, from the air swirling around an aircraft wing to molten metal in industrial processes or even blood flow in the human body. What makes this achievement so monumental is not only the depth of the physical insight, but also its timelessness — the equations were formulated in the 19th century, yet they are still the foundation of modern simulations, engineering design, and even AI-driven reduced-order models today.
What direct or indirect relevance does your current research have for the world?
In the simplest way I would say that my research is about making computer simulations of fluids — like hot gases, molten metals, or even water with lots of movement — much faster by using machine learning. Why does this matter? Because these kinds of flows appear everywhere: in making steel and glass, in producing energy, even in climate and medical research.
Stay tuned for Part 2, where Amra reflects on the life of a researcher, ways to improve the situation of women in science, and her favorite discoveries in Münster!