Physics Department · Universidad de los Andes
Jaime Forero-Romero
Associate Professor of Physics
I study the large-scale structure of the Universe: the cosmic web of filaments, walls, and voids that connects every galaxy, using astronomical surveys, cosmological simulations, and machine learning. I've been a member of the Dark Energy Spectroscopic Instrument (DESI) collaboration since 2014 and have led a computational cosmology research group at Universidad de los Andes since 2012.
About
My work spans three connected goals: understanding how the cosmic web shapes galaxies, using the cosmic web to test cosmology, and building AI/ML methods that serve both. I currently hold an Associate Professor position (with tenure) in the Physics Department at Universidad de los Andes in Bogotá, Colombia, where I've taught and mentored students since 2012.
Before returning to Colombia, I was a Gruber Fellow in the Astronomy Department at UC Berkeley (2011-2012) and a postdoctoral fellow at the Leibniz Institute for Astrophysics Potsdam (AIP) in Germany (2007-2011). I received my PhD in Physics from École Normale Supérieure de Lyon (Université de Lyon), France, in 2007, after a Magistère in Physics at École Normale Supérieure, Paris, and undergraduate studies at the Instituto Balseiro (Argentina) and Universidad Nacional de Colombia.
Research
My research is organized around the cosmic web, used in three connected ways: as a tool to understand galaxies, as a tool for cosmology, and as a playground for state-of-the-art AI and machine learning methods that benefit both.
The cosmic web and large-scale structure
In 2009 I introduced a dynamical classification of the cosmic web that splits space into voids, sheets, filaments, and knots by counting the eigenvalues of the gravitational deformation tensor at each point in a simulation. It has since become a standard reference framework for connecting cosmic web environment to halo and galaxy properties. That approach needs a continuous, gridded density field, easy in a simulation box but hard in a real spectroscopic survey with a mask, a selection function, and gaps. In 2025 my group closed that gap with ASTRA (Algorithm for Stochastic Topological RAnking), which classifies galaxies by comparing their observed positions against a random catalog, the same trick redshift-survey clustering codes use to handle survey geometry. ASTRA is on track to become a core tool for cosmic web science over the coming years of survey data.
The Dark Energy Spectroscopic Instrument
My most substantial ongoing commitment has been to DESI, the Stage-IV spectroscopic survey mapping the Universe out to z≈4 from the Mayall Telescope at Kitt Peak. Since joining in 2014 I've contributed to fiber assignment, target selection, and data quality assurance, work recognized through DESI Builder status. Concretely this has included building and quality-checking several of DESI's main tracer samples, supporting production of the alternate fiber-assignment catalogs needed for robust covariance matrices, and applying unsupervised machine learning (UMAP) to flag instrumental outliers in nightly spectroscopic data. Within the collaboration I've also led focused analyses, including the first detection of the large-scale tidal field through the alignment of galaxy multiplets in DESI data. I now co-chair DESI's Galaxy Environment topical group, producing key papers on how environment shapes galaxy properties.
Machine learning for cosmology
My group applies deep learning across several problems in cosmology: recovering cosmic web classifications directly from galaxy positions with graph-based features, reconstructing the cosmic velocity field from galaxy positions, estimating cosmological parameters directly from large-scale structure data instead of from compressed summary statistics, and classifying image sequences of astronomical transients with deep neural networks. On the operational side, I use unsupervised learning to automate quality assurance for DESI's nightly spectroscopic data.
Looking ahead
Going forward, my research keeps working on three fronts at once: understanding how the cosmic web shapes galaxies, developing it further as a clustering method that goes beyond standard two-point statistics, and using AI/ML to make both work at survey scale. This includes continuing to co-lead DESI's Galaxy Environment key papers; contributing to DESI's Alternative Clustering Methods effort, which covers higher-order statistics, void statistics, and other nonlinear measures of the galaxy distribution; an ongoing collaboration connecting cosmic web environment to galaxy gas content; and moving from compressed summary statistics toward full simulation-based inference using AI/ML. I'm also building on a decade of regional leadership in Latin American astronomy, including the Colombian Astronomical Community and the Andean Regional Office of Astronomy for Development, to grow a research group of PhD students and postdocs with a distinctive pipeline of talent from across the region.
Publications
Selected publications
Five publications that best represent my research, in narrative order: a foundational method, its modern successor, survey infrastructure, a flagship result, and a method applied back to survey data.
-
Forero-Romero, J. E., Hoffman, Y., Gottlöber, S., Klypin, A., Yepes, G. (2009). A dynamical classification of the cosmic web. MNRAS, 396, 1815-1824. doi:10.1111/j.1365-2966.2009.14885.x
Introduced a dynamical classification of the cosmic web into voids, sheets, filaments, and knots by counting the eigenvalues of the gravitational deformation tensor above a threshold at each point in a simulation. It remains a standard framework for connecting cosmic web environment to halo and galaxy properties.
-
Forero-Romero, J. E., Palomino, A., Gómez-Cortés, F. L., Li, X. (2025). Cosmic web classification through stochastic topological ranking. RAS Techniques and Instruments, 4, rzaf032. doi:10.1093/rasti/rzaf032
ASTRA classifies galaxies into voids, sheets, filaments, and knots by comparing observed positions against a random catalog, avoiding the density-field interpolation earlier methods required and enabling robust classification even where galaxies sparsely sample a survey.
-
Abdul Karim, M., et al. (200 authors) (2026). Data Release 1 of the Dark Energy Spectroscopic Instrument. AJ, 171, 285. doi:10.3847/1538-3881/ae4c43
DESI's first major public data release: high-confidence redshifts for 18.7 million objects spanning z=0 to z≈4, the largest extragalactic redshift sample ever assembled.
-
Adame, A. G., et al. (200 authors) (2025). DESI 2024 VI: cosmological constraints from the measurements of baryon acoustic oscillations. JCAP, 2025, 021. doi:10.1088/1475-7516/2025/02/021
DESI's headline BAO cosmology paper, combined with CMB and supernova data to find a preference for time-varying dark energy over the standard cosmological constant, one of the most widely discussed cosmological results of the decade.
-
Lamman, C., Eisenstein, D., Forero-Romero, J. E., et al. (39 authors) (2024). Detection of the large-scale tidal field with galaxy multiplet alignment in the DESI Y1 spectroscopic survey. MNRAS, 534, 3540-3551. doi:10.1093/mnras/stae2290
Detects intrinsic alignment of small galaxy groups with the large-scale tidal field out to 100 h⁻¹ Mpc, establishing multiplet alignment as a promising new probe of cosmological parameters and the galaxy-halo connection.
Teaching
I see knowledge as three interconnected dimensions: being part of a community, sharing a body of validated knowledge within it, and sharing its practices. Since 2012 I have taught across all levels at Universidad de los Andes, almost every semester, building all three dimensions into my courses through group work and field-specific language and problems, since physics and astronomy are best taught close to real data and real problems, not only textbook exercises.
I collect structured feedback at mid-semester and again at the end of each course, using simple start/stop/keep principles. My teaching evaluations show a clear improving trend, now averaging 0.5 standard deviations above the university-wide average across all five evaluated dimensions.
My approach to supervision is to give students a real, well-scoped piece of an active research problem, usually built on DESI data or my cosmic web methods, with regular structured feedback. Since 2012 I have supervised 19 undergraduate theses, 3 Master's theses, 1 completed PhD thesis, and 3 postdoctoral researchers.
Curriculum vitae
Academic appointments
- 2015-present Associate Professor (with tenure), Physics Department, Universidad de los Andes, Bogotá
- 2012-2015 Assistant Professor, Physics Department, Universidad de los Andes, Bogotá
- 2011-2012 Gruber Fellow, Astronomy Department, UC Berkeley
- 2007-2011 Postdoctoral Fellow, Leibniz Institute for Astrophysics Potsdam (AIP), Germany
Education
- 2005-2007 PhD in Physics, École Normale Supérieure de Lyon (Université de Lyon), France
- 2003-2005 Magistère Interuniversitaire de Physique, École Normale Supérieure, Paris (Master, Sciences de l'Univers, Astronomie et Astrophysique; Maîtrise de Physique, Université Pierre et Marie Curie)
- 2001-2003 Undergraduate studies in Physics, Instituto Balseiro, Bariloche, Argentina
- 1999-2001 Undergraduate studies in Physics, Universidad Nacional de Colombia, Bogotá
Awards & honors
- 2011-2012 Gruber Fellowship, UC Berkeley
Selected leadership & service
- 2024-present Board Member, Colombian Astronomical Community (AstroCO)
- 2023-present DESI Builder
- 2017-2022 Colombian Coordinator, Latin American Chinese European Galaxy Formation Network (LACEGAL), EU Horizon 2020
- 2016 Co-organizer, XV Latin American IAU Regional Assembly
- 2015-2020 Founder & Coordinator, Andean Regional Office of Astronomy for Development (IAU)
- 2020-present CINFONIA (Centro de Investigación y Formación en Inteligencia Artificial), Universidad de los Andes
- 2010-present Referee, The Astrophysical Journal and Monthly Notices of the Royal Astronomical Society