Dr. Abdurro’uf
I turn pixels into physics. I map the internal structure of galaxies across 12 billion years of cosmic time and connect telescope data to simulations through forward modeling and machine-learning inference.
Bloomington, IN, USA
Bridging observations, simulations, and machine learning
Galaxies are not points. The processes that build and quench them, including gas inflow, compaction, stellar and black-hole feedback, mergers, and environment, act in specific places within galaxies and on specific timescales.
I am an extragalactic astrophysicist working at the intersection of astrophysics and data science. I map the spatially resolved physical properties of galaxies with imaging and spectroscopic data from various telescopes (including HST, JWST, and Euclid), and compare them with cosmological hydrodynamical simulations. I build the open-source tools that link the two: piXedfit for spatially resolved SED fitting, together with its successor piXedfit2 (in development), which combines traditional Bayesian SED fitting and machine-learning inference in one modular, interactive framework; GalSyn for forward-modeling simulations into realistic observations; and GalSyn-SBI, a set of simulation-based inference (SBI) networks that recover global and resolved galaxy properties in under a second. These tools are built for the coming flood of data from Euclid, Roman, and Rubin/LSST.
I was PI of JWST Cycle 2 program GO 4246 on the study of a merging galaxy at the cosmic dawn epoch, and I am an active member of the Euclid Consortium and the Roman Science Collaboration, and a co-investigator on more than a dozen JWST programs. Before Indiana, I was a postdoctoral fellow at Johns Hopkins University and ASIAA (Taiwan), and I earned my PhD at Tohoku University (Japan).
Observe
Resolved maps of galaxies from z ~ 0 to the cosmic dawn epoch with JWST, HST, and Euclid.
Simulate
Forward-model TNG, EAGLE/FLARES, and SIMBA into realistic mock data.
Infer
Train calibrated SBI networks on mocks and apply them to millions of galaxies.
From pixels to physics
One question drives my program: how do galaxies assemble their internal structures and shut down their star formation across 12 billion years, and which physical processes set the routes they take? I answer it with a closed loop between simulations and survey data.
- Simulations provide labeled ground truth: galaxies whose full history is known.
- GalSyn forward-models them into realistic synthetic data (HST, JWST, Euclid, Roman, and Rubin).
- SBI networks (GalSyn-SBI) trained on these mocks infer global and resolved properties in <1 s per galaxy.
- Applied to real surveys, they yield resolved maps and SFHs for 105–107 galaxies.
- These maps test hypotheses about quenching routes and structural assembly.
- The results constrain the physics in the next generation of simulations.
Research themes
THEME 01The Early Universe
The first galaxies and their star clusters seen with JWST: ISM conditions at z > 10 and parsec-scale clumps in strongly lensed galaxies.
THEME 02Structural Evolution of Galaxies
How galaxies build bulges and disks and shut down star formation, inside-out, mapped pixel by pixel from the cosmic dawn epoch to today.
THEME 03Forward Modeling of Simulations
GalSyn turns hydrodynamical simulations into realistic telescope data with known ground truth, so observations and theory meet on equal footing.
THEME 04Machine-Learning Inference
GalSyn-SBI: simulation-based inference that recovers global and resolved property maps and star-formation histories in under a second per galaxy.
Open-source tools for the survey era
Public, documented Python packages for resolved galaxy analysis and simulation forward modeling.
GalSyn
Forward-models hydrodynamical simulations into realistic imaging and IFS data, together with the ground-truth property maps behind them.
piXedfit
An end-to-end package for spatially resolved SED fitting: image processing, PSF matching, imaging–IFS combination, pixel binning, and Bayesian SED fitting.
piXedfit2
A modular, pythonic rewrite of piXedfit for interactive resolved SED analysis from survey mosaics, with five inference backends: RDSPS, MCMC, Dense Basis, Bagpipes, and SBI.
GalSyn-SBI
Simulation-based inference that turns multiband images into calibrated maps of stellar mass, SFR, age, metallicity, and dust in under a second, using a U-Net and normalizing flows trained on GalSyn mocks.