Interview with Malgorzata Siudek

Postdoctoral researcher at Institut de Física d’Altes Energies

 

What is your current role within the ESA Euclid mission?

I am a postdoctoral researcher working within the Euclid Consortium, where my role focuses on developing and applying machine learning methods to analyse Euclid's data. I became actively involved shortly before the internal release of Euclid's first data (Q1) in October 2024. Since then, I've contributed to several working groups and led one of the first studies using Q1 data, demonstrating how AI models can classify galaxies and estimate their properties using Euclid's high-resolution images. While I still consider myself early in my Euclid journey, this mission offers an incredible opportunity to explore the universe at a scale and depth we've never had before, and I'm excited to become more involved in the years ahead.

In parallel, I'm also a member of the DESI survey team—the Dark Energy Spectroscopic Instrument—which I see as highly complementary to Euclid. While Euclid delivers high-resolution imaging and near-infrared spectroscopy from space, DESI provides high-resolution optical spectroscopy from the ground. Although the overlap between the two surveys is currently limited, I believe their synergy will be powerful in the future. Being involved in both allows me to study galaxy evolution from different perspectives, and more multi-wavelength approach.

Have there been any unexpected findings or surprises in the ESA Euclid mission so far?

Yes, I think Euclid is following a similar trajectory to the James Webb Space Telescope (JWST) - starting with strong technical validation and quickly delivering groundbreaking science. It's already clear with Q1 data, that Euclid is already challenging our assumptions. We expected high-quality data, but the volume and variety of novel discoveries in such an early phase are truly above our expectations.

In particular, Euclid has revealed a bunch of hidden populations: strong gravitational lensing systems, faint dwarf galaxies, active galactic nuclei (AGN), and even rarer "little red dots", which are small, compact, and extremely red galaxies, previously detected by JWST, that challenge our understanding of early galaxy formation. This is only the beginning, and I truly believe many more surprises lie ahead.

What part of the Q1 data release and your own research are you most excited about?

What excites me most is how Euclid is transforming our ability to study a very large number of galaxies at once, and how to do that with smarter tools. Galaxies come in many shapes, colours, and distances, and by studying them, we can trace the history of the Universe and even predict its future. Euclid is the perfect tool to study the accelerated expansion of the Universe, but with this great number of observations, we need smart tools to analyse them quickly, as we cannot go through them one by one.

That's where our work comes in. We developed AstroPT, a powerful AI foundation mode. What makes AstroPT unique is that it is a multi-modal model that learns from galaxy images and light patterns (spectral energy distributions). Previously, we could only look at images or spectra separately, combining multi-wavelength information is just being developed. AstroPT model can be tuned for specific tasks, and for that, it can still work with very little training data. We only need a few - at least few in the Euclid perspective - galaxies with known shapes or redshifts to train the model, and it can classify galaxy shapes, estimate redshifts, and spot rare objects very well. This is very important in the era of Euclid-like surveys, which reach deeper and further than before, allowing us to find hidden objects in an automated way.

Based on the knowledge you have now from the Q1 data release, what are your main expectations from Euclid in the future?

Euclid will allow us to understand the large-scale structure of the Universe, dark matter, and dark energy. But I am mostly interested in how Euclid can contribute to our understanding of Galaxy evolution. I think that Euclid will provide us with large data sets that will allow us to better study this galaxy evolution, especially by discovering peculiar, rare galaxies that were not visible to us with other missions. But I also believe it is a platform for developing foundation models, which is a step forward for AI-powered astronomy. With these large datasets delivered by Euclid, we can build models that help us discover the universe's secrets faster and more reliably. Just as models like ChatGPT have changed how we interact with language, I expect that Euclid will help us build models that change how we explore the Universe: faster, more flexible, and with greater insight.

Do you have any advice for people who would like to follow a similar career path or to use Euclid data?

My advice is: don't be intimidated. Getting involved with Euclid data is more accessible than people might think. Most researchers I know are incredibly open and passionate about sharing knowledge. We regularly give outreach talks and are always happy to collaborate or answer questions.
ESA has done a great job of making the data public and easy to access, so anyone can start exploring. Play around with it, see what interests you, and if you need help, do not hesitate to reach out to researchers or join a community event. I believe that communication is key to science, so just collaborate and reach out.
I consider myself very lucky to have made my passion my work, and I always encourage others to follow the same path. If you are excited about the Universe, go for it. Believe in yourself, have the courage to get involved, and be ready to put in the effort - because making a real contribution takes study, persistence, and hard work. But if you love it, it is absolutely worth it.

 

 

For further details on Malgorzata Siudek's work on a multi-modal foundation model used to explore galaxy properties, please refer to the following scientific paper submitted to the arXiv:

For more information about the Euclid Q1 release, visit the ESA press release: