Ali Vaziri

Developing software for AI and GPU-based infrastructures with a focus on performance, scalability, and reliability, while optimizing the cloud platform for AI training, inferencing, and other intensive workloads.

Developed numerical methods for compute-intensive HPC applications and conducted algorithmic research. Designed and implemented infrastructure for testing HPC applications and parallel I/O. Developed physics-informed neural networks (PINNs) for solving PDEs and accelerated convolutional neural networks (CNNs) using Ray/Horovod.

Developed computational methods for the ExxonMobil iRMS software including optimization methods for conditioning reservoir models to seismic/well log data, algorithms for wavelet-based seismic image processing, computational geometry and model sharing (RESQML data standards), rock physics and forward seismic modeling.

Developed cross-platform distributed hybrid (MPI+OpenMP/CUDA) sparse direct multi-frontal solvers for large scale problems simulated by MSC Apex and MSC Nastran softwares in the High Performance Computing (HPC) and Scalable Physics Framework (SPF) Departments.

Developed the Discontinuous Petrov-Galerkin methodology under supervision of Professor Leszek F. Demkowicz: Developed high-order Polygonal Discontinuous Petrov-Galerkin (PolyDPG) methods Developed goal-oriented adaptive mesh refinement strategies for non-symmetric functional settings Developed perfectly matched layers for Petrov-Galerkin formulations

Developed rapid convergent forward modeling and inversion algorithms for biomedical imaging, near surface characterization and nondestructive testing problems. Developed an accurate and stable absorbing boundary conditions for elastic waveguides. Developed a fast domain decomposition based preconditioner for solving large scale wave propagation problems.

Developed a new distortion-tolerant finite element for large deformation analyses. Developed meshfree methods for: elastodynamics, elastic fracture, acoustics and heat propagation.