Phám, Thanh Son2026-01-022026-01-020956-540XORCID:/0000-0002-9057-4416/work/193550755https://hdl.handle.net/1885/733802811The misalignment of the observation and predicted waveforms in regional moment tensor inversion is mainly due to seismic models' incomplete representation of the Earth's heterogeneities. Current moment tensor inversion techniques, allowing station-specific time-shifts to account for the model error, are computationally expensive. Here, we propose a gradient-based method to jointly invert moment-tensor parameters, centroid depth and unknown station-specific time-shifts utilizing the modern functionalities in deep learning frameworks. A misfit function between predicted synthetic and time-shifted observed seismograms is defined in the spectral domain, which is differentiable to all unknowns. The inverse problem is solved by minimizing the misfit function with a gradient descent algorithm. The method's feasibility, robustness and scalability are demonstrated using synthetic experiments and real earthquake data in the Long Valley Caldera, California. This work presents an example of fresh opportunities to apply advanced computational infrastructures developed in deep learning to geophysical problems.This work greatly benefited from discussion with Jinyin Hu and Hrvoje Tkal\u010Di\u0107 through ongoing research on MT inversion. The Air Force Research Laboratory's grant, contract number FA9453-20-C-0072, supported the author's post doc at The Australian National University. He also acknowledges financial support from the Australian Research Council through a Discovery Early Career Researcher Award, project DE230100025. This research was undertaken with the assistance of resources from the National Computational Infrastructure (NCI Australia), an NCRIS enabled capacity supported by the Australian Government. The author thanks Editor Carl Tape, Ji\u0159\u00ED Vack\u00E1\u0159 and an anonymous reviewer for constructive review comments, which significantly improve the quality of the this paper.11en© 2024 The Author(s). Published by Oxford University Press on behalf of The Royal Astronomical Society.Computational seismologyjoint inversionMoment tensorOptimizationTheoretical seismology202410.1093/gji/ggae18885196097616