Nina Gottschling investigates how mathematical constraints shape reliability in AI, scientific computing

Published: February 25, 2026

Updated: February 25, 2026

Nina Gottschling, a Wigner Fellow at the Department of Energy's Oak Ridge National Laboratory whose research spans uncertainty quantification, inverse problems and photonoic quantum computing, began her path into mathematics and scientific computing with an early fascination for physics, logic and philosophy. As an undergraduate studying physics and philosophy at Ludwig Maximilians University in Munich, Germany, she found herself captivated by the precision and expressive power of mathematics.

"I really loved how you could express yourself so precisely in mathematics," she said.

That curiosity carried her from a dual bachelor's degree into an elite master's program in theoretical and mathematical physics, and ultimately to the University of Cambridge - thanks to a last-minute doctoral application submitted just one week before the deadline, "I applied… and suddenly I was moving to Cambridge to start a fully funded doctorate in mathematics," she said.

At Cambridge, Gottschling took on a scientifically fundamental research question: Why do deep neural networks for inverse problems hallucinate and become unstable - and how can we mathematically understand when and why this occurs? Her doctoral work explored the existence, accuracy and stability of approximate decoders for ill-posed inverse problems, producing theoretical insights that informed practical applications across microscopy, satellite imaging and scientific computing.

A researcher drawn to developing accuracy bounds for inverse problems and scientific computing

The difficulty of the questions she pursued did not deter her - though they tested her. Early in her doctorate, she and her collaborators developed a conjecture that took years to prove. "It was depressing at times. I almost started studying law because I wasn't sure it would ever work," she added.

But persistence proved worthwhile. Her research began uncovering accuracy bounds for real-world inverse problems, laying the groundwork for later applications in multispectral satellite super-resolution and localization for microscopy - areas where she found meaning in the ability to compare different state-of-the-art methods for solving inverse problems on mathematically grounded terms.

Following her doctorate, Gottschling returned to Germany and joined the German Aerospace Center (DLR), where she worked across a broad range of research directions: uncertainty quantification for satellite data, photonic quantum computing and AI analysis for Earth observation. "I had the opportunity to do anything I wanted… so I did the things I thought were fun," she said.

Her contributions at DLR extended beyond theory. She collaborated on applications including flood crisis management, real-world adversarial attack detection and defense for object-detection systems, and uncertainty-aware function approximation on integrated photonic circuits. These projects demonstrated how her mathematical insights could drive impactful engineering solutions across secure sensing and emerging quantum technologies.

Finding a home at ORNL to advance AI reliability, quantum systems and mathematical computing

A turning point came during her participation at the Lindau Nobel Laureate Meeting in Physics. Inspired by lectures and conversations about the future of AI, energy use and scientific computing, she began considering how her interests in quantum systems, machine learning and inverse problems could merge.

"I saw that these fields seem to be important and that there are open problems - we don't understand what this black box of AI actually does," she said. That realization sparked an idea: combining quantum neural network implementations with mathematical frameworks for accuracy and efficiency to potentially reduce the energy footprint of modern AI systems.

Encouraged by friends - and her now-fiancé - she applied to ORNL's Wigner Fellowship. "I thought it would be really hard to get into… and somehow it worked out," she said.

Today, as a Wigner Fellow in ORNL's Computational Sciences and Engineering Division, Gottschling is pursuing accuracy bounds and mathematical frameworks for scientific computing systems - research that lies at the intersection of rigorous mathematics, computational simulation and experimental collaboration.

"I see my work in three layers," she added. "Mathematics is about 70 percent of what I do. Then comes simulating or computing based on those structures. And finally, collaborating with people to test those ideas in experiments."

Her current work includes developing new mathematical tools that help quantify how reliable complex computational models can be, even as systems grow larger and more interconnected. These efforts aim to better understand fundamental limits on accuracy in scientific computing and machine learning. She is also in the process of writing her Laboratory Directed Research and Development proposal. Nina explains that a central motivation in her research is helping AI engineers understand, before training any model, how good a model can possibly be on a given dataset and whether that dataset may need improvement. She sees this kind of mathematical guidance as valuable for reducing unnecessary computation and clarifying what is achievable in real-world scientific applications.

A philosophy of precision and logic guiding advances in inverse problem theory

Gottschling's approach to research is shaped by her early love of logic and philosophy of science. She begins by learning as much as she can about a problem before attempting to structure it.

"I try not to judge anything at first because you might miss something if you have preconceptions," she said. From there, she turns to mathematics - the language she trusts most - to build frameworks for understanding and solving the problem at hand.

She emphasizes that mathematics must remain grounded in reality. "You can prove anything if you make the right assumptions - but the key is making assumptions based on what engineers actually do," she said. "Otherwise it's like putting salt in your coffee and hoping it will be sweet."

This philosophy shapes her long-term scientific vision as well. Five to ten years from now, she hopes to contribute to bridging the gap between theoretical computing frameworks - using tools such as category theory - and practical computing systems. She would also like to contribute to understanding AI hallucinations in large language models. She also hopes to explore how accuracy bounds behave for numerical solutions of the Navier-Stokes equations, a direction she finds "very interesting," particularly because it raises fundamental questions about whether solutions to the Navier-Stokes equations are unique; and if they are not, how scientists should approach that uncertainty.

Early impressions of ORNL's interdisciplinary environment for AI, scientific computing

Though she joined ORNL only recently, Gottschling has already found the environment energizing. She points to the lab's breadth of scientific expertise and the close coexistence of theory, simulation and experimental science as especially compelling.

"Whenever you think of something mathematically or simulate it, it's always good to test it. Otherwise, you might be doing something quite useless."

She has been especially impressed by ORNL's experimental physicists - including one researcher who was testing and validating new quantum hardware without simulation tools. "I thought, don't you need someone to help you do that? You could, but you don't have the time. Simulation would make your work so much easier," she added.

She credits Vladimir Protopopescu Computational Sciences and Engineering Division's chief scientist, with helping her navigate the lab's structure during her first weeks on site.

A researcher who finds balance through movement, problem-solving and nature

Outside the lab, Gottschling is almost always moving. She loves climbing, running and mountain biking - activities she describes not just as exercise but as mental puzzles. "Being able to put yourself on the rock and figure out the puzzle of how to move - that's what I love," she said.

She is also an avid sailor and a fan of weather models, a hobby that hints at the same analytical curiosity that drives her scientific work. She shares her home with her partner and their cat and says she cherishes evenings spent solving science questions together just for fun.

Tennessee, she says, has already delighted her with its outdoor community - especially its well-maintained mountain biking trails. She looks forward to exploring more of the region.

A fellow focused on purpose and advancing the future of scientific computing

Although she has only just begun her Wigner Fellowship, Gottschling feels both honored and motivated by the opportunity. "With great opportunities come big responsibilities," she said. "I aim to deliver on what I promise and hope I can achieve real benefit for engineers."

Her advice to future fellows is simple: "Don't hesitate to apply, even if your h-index isn't amazing. If you have an idea that might be valuable to DOE, just try."

If there is one thing she hopes readers understand about her work, it is that mathematics is most powerful when connected directly to engineering practice - and that scientific curiosity, even when it takes unexpected turns, can lead to profoundly meaningful discoveries.

"I'm happy looking at scientific problems," she said. "And even if it doesn't always look directed, sometimes coincidences end up in something really cool."

ORNL's Distinguished Staff Fellowship program aims to cultivate future scientific leaders by providing dedicated mentors, world-leading scientific resources and enriching research opportunities at a national laboratory. Fellowships are awarded to outstanding early-career scientists and engineers who demonstrate success within their academic, professional and technical areas. Fellowships are awarded for fundamental, experimental and computational sciences in a wide range of science areas. Factsheets about the lab's fellows are available here.

UT-Battelle manages ORNL for DOE's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE's Office of Science is working to address some of the most pressing challenges of our time. For more information, visit energy.gov/science. - Neil Gillette

Media Contact

Elicia M Ferrer , Executive Communications Specialist , 865.341.3240 | [email protected]