Postdoctoral Fellow
Research group |?Cosmology and Extragalactic Astrophysics
Main supervisor |?Hans Arnold Winther
Co-supervisor |?-
Affiliation |?Institute of Theoretical Astrophysics
Contact |?andriust@uio.no
Short bio
I am a Marie Sk?odowska-Curie Actions (MSCA) DSTrain Fellow at the Institute of Theoretical Astrophysics. I am a cosmologist focusing on developing machine learning, numerical, and simulation tools to search for new physics. I am particularly interested in applying recent breakthroughs in artificial intelligence (AI) to a range of problems in cosmology.
Position history:
? Marie Sk?odowska-Curie Actions (MSCA) DSTrain Fellow: Institute of Theoretical Astrophysics, University of Oslo, Norway (2025–present)
? Postdoctoral Fellow: split between Case Western Reserve University, USA (as a Richard S. Morrison Fellow) and the Institute of Theoretical Physics (IFT), UAM-CSIC, Spain (as a Postdoctoral Researcher) (2022–2025)
? Postdoctoral Research Associate: University of Nottingham, United Kingdom (2020–2022)
? Data Intensive Science (DISCnet) PhD in Cosmology: Institute of Cosmology and Gravitation (ICG), University of Portsmouth, United Kingdom (2017–2020)
Research interests and hobbies
While the standard model of cosmology has been highly successful, recent DESI findings suggesting a time-varying nature of dark energy, together with the H? and other tensions, indicate that the ΛCDM model may need to be extended, for example by modifying Einstein’s general relativity. A major success of the standard model is its explanation of the Cosmic Microwave Background (CMB). However, several observed CMB anomalies, such as the low quadrupole power, alignment of low multipoles (the “Axis of Evil”), hemispherical power asymmetry, and the Cold Spot, suggest that our understanding may be incomplete. While some anomalies could be statistical flukes or systematic effects, collectively they remain difficult to explain and may hint at new physics or the Universe having a non-trivial cosmic topology.
My recent work explores AI, numerical, and simulation techniques to study observational signatures of modified gravity theories. Together with collaborators, we have developed SELCIE, a numerical code for solving field equations in modified gravity theories, and applied it to optimise atom interferometry experiments. Using SELCIE and genetic algorithms, we have also optimised astrophysical tests of gravity, identifying density profiles that maximise the potential fifth-force signal in galaxy clusters and cosmic voids. These results help refine observational strategies for Euclid and other next-generation surveys.
On the CMB side, I develop AI-based methods to search for evidence of new physics. As part of the COMPACT Collaboration, I have designed machine learning approaches, such as spherical graph convolutional neural networks, to detect signatures of non-trivial cosmic topology in the CMB data. I am also interested in theoretical explanations of the CMB anomalies.
Outside of work, I enjoy cycling, hiking, and wild camping. I look forward to exploring Norway’s nature, improving my cross-country and downhill skiing, and seeing the Aurora Borealis for the first time.
DSTrain project
In Search of Physics Beyond ΛCDM with Stage-IV Cosmological Surveys and Machine Learning
As a DSTrain Fellow, I hope to use some of the outlined techniques to optimise next-generation astrophysical, space-based, and laboratory tests of gravity, for example by identifying the astrophysical systems that are most promising for detecting the effects of modified gravity and dark matter. I also aim to apply the newest advances in generative machine learning as tools for data analysis and novel data generation, which will be particularly important for emulating the data required by Euclid and other surveys. In addition, I am interested in applying recent advances in large language models (LLMs) as tools in cosmology. New approaches, such as agentic AI systems, show great promise in general reasoning, code generation, and iterative improvement through fitting scientific data. Methods like these will open many exciting avenues for AI applications in cosmology.
?
Publications
DSTrain publications
ORCID ID: 0000-0003-2907-4575
Previous publications
?
?
?
?
?
?