Research
My research centers on developing and applying deep learning algorithms for astrophysical and cosmological observations, with a focus on image reconstruction and analysis. In particular I work with radio interferometric observations of supermassive black holes and relativistic jets and observations of the cosmic microwave background.
A deep learning algorithm for VLBI imaging
kine is a dynamic imaging algorithm for VLBI observations, based on neural fields. kine reconstructs videos of sources with intra-day variability as well as videos of slowly varying sources from multi-epoch monitoring.
Turbulent flow in the relativistic jet of 3C 345 from neural video reconstructions of interferometric data
M. Foschi*, B. Zhao*, A. Fuentes*, K. L. Bouman, J. L. Gómez, A. Levis.
Under review, (2025)
* Equal author contribution
Dynamics of the SgrA* black hole
SgrA* is a supermassive black hole located at the center of the Milky Way. The kine imaging algorithm is being applied to observations of SgrA* by the Event Horizon Telescope to obtain the first horizon-scale video of the variable emission produced from the accretion disk around the black hole.
Validation of horizon-scale Sagittarius A* video reconstruction with kine
A. Fuentes*, M. Foschi*, R. Dahale, J. L. Gómez, A. Levis, N. S. Conroy, K. L. Bouman, EHT Collaboration
In preparation (2026)
* Equal author contribution
Evolution of the 3C 84 relativistic jet
Forward imaging methods can improve the resolution of interferometric images of relativistic jets, enabling the measurement of instantaneous kinematic properties of the jet evolution.
Evolution, speed, and precession of the parsec-scale jet in the 3C 84 radio galaxy
M. Foschi*, J. L. Gómez, A. Fuentes, I. Cho, A. P. Marsher, S. Jorstad
Astronomy & Astrophysics, 696, A17 (2025)
CMB Galactic Foregrounds
FORSE+ is a Python package that produces non-Gaussian Galactic thermal dust emission maps at arcminute resolution with the possibility of generating random realizations of the small scales.
ForSE+: Simulating non-Gaussian CMB foregrounds at 3 arcminutes in a stochastic way
based on a generative adversarial network
J. Yao, N. Krachmalnicoff, M. Foschi, G. Puglisi, C. Baccigalupi
Astronomy & Astrophysics, 686, A290 (2024)
