NESAP Postdocs
Since introducing the NERSC Exascale Scientific Applications Program (NESAP) in 2015, NERSC has hired a continuous contingent of postdoctoral fellows and placed them with NESAP code optimization teams. The postdocs are working in multidisciplinary teams composed of computer scientists, applied mathematicians, domain scientists and performance optimization experts that are helping NERSC users transition codes to the Cori system.
Name | Hire Date | Ph.D. Institution | Ph.D. Discipline | Project / Contributions |
Soham Ghosh | 2020 | Florida State University | Physics | Developing and accelerating NESAP for simulating massively parallel software for computing electronic structures within many-body perturbation theory. The work involves porting and scaling both screening and correlation methods on the GPU as well as making the routines more portable between different accelerating hardware. |
Dhruva Kulkarni | 2020 | Clemson University | Physics | Developing and accelerating NESAP for Simulations application WDMApp. Evaluating and optimizing the performance of various kernels and solvers in WDMApp across different hardware (NVIDIA, AMD, Intel) and software (Kokkos, OpenMP acceleration) stacks. |
Nestor Demeure | 2021 | University Paris Saclay | CS and Applied Math | Adding GPU capabilities to the TOAST Astrophysics application stack which is used to study the Cosmic Microwave Background. This implied updating the benchmarking infrastructure of the project. Adding support or GPU libraries including CuBLAS and CuFFT. Implementing and optimizing GPU kernels in both JAX and OpenMP target offload in order to compare both technologies. |
Felix Wittmer | 2021 | Zurich U. | Physics | Acceleration of the EXAFEL (LCLS) application stack. The CCTBX project currently relies on Nvidia/CUDA to achieve GPU acceleration. To be more vendor independent, the project involves porting the nanoBragg code from CCTBX to Kokkos and evaluate the performance on Perlmutter, Frontier and Aurora. |
Nick Tyler | 2021 | U. of SC | Physics | Workflow and GPU optimization of JGI (genomics) pipelines |
Shashank Subramanian | 2021 | UT Austin | CS | AI based acceleration of computational fluid dynamics and weather modeling NeuWCast. |
Lipi Gupta | 2021 | U. of Chicago | Physics | Acceleration of beamline science workflows from the ALS to NERSC |
Vinicius Mikuni | 2021 | Implementation and optimization of machine Learning algorithms in High Energy Physics. | ||
Richard Barnes | 2021 | UC Berkeley | Geophysics | Applying “depression hierarchy” algorithms, learned indexes and AI-based algorithms (Deep neural networks) to important problems in geoscience, electric public transportation, connectomics and Genomics. |
Name | Term | Ph.D. Institution | Ph.D. Discipline | After PostDoc | Project / Contributions |
Jihan Kim | 2009 - 2013 | Computational Physics | Korea Advanced Institute of Science and Technology | Developed a GPU code to accelerate screening of microporous materials to absorb carbon dioxide gas from power plant flue gases. | |
Robert Preissl | 2010 - 2011 | Computer Science | IBM; Ticketfly; Kitty Hawk | Worked with physicists at the Princeton Plasma Physics Laboratory (PPPL) to implement PGAS and hybrid programming solutions for highly scalable particle-in-cell codes | |
Xuefei (Rebecca) Yuan | 2010 - 2012 | Applied Mathematics | Bank of America; Wells Fargo | Improved hybrid linear software routines based on the Schur complement method in collaboratoin LBNL CRD. | |
Wangyi (Bobby) Liu | 2010 - 2013 | Applied Mathematics | Improved the ALE-AMR code and developed scalable solutions for new physics models for Heavy-Ion science. | ||
Brian Austin | 2010 - 2011 | Chemistry | LBNL NERSC | Developed simulations for next-generation light sources based on x-ray free electron lasers | |
Kjiersten Fagnan | 2010 - 2012 | Applied Mathematics | LBNL NERSC; LBNL JGI | Developed numerical simulations of carbon sequestration and porous media flow. | |
Filipe Maia | 2010 - 2012 | Physics (Molecular Biophysics) | LBNL ALS; Uppsala | Accelerated biological imaging codes and Earth Sciences Division and geophysical imaging codes with GPUs | |
Praveen Narayanan | 2010 - 2012 | Mechanical Engineering | NVIDIA; Ford | Completed performance characterization and benchmarking of several parallel applications running at NERSC, in collaboration with code teams at several DOE facilities. | |
Christos Kavouklis | 2011- 2015 | Engineering Mechanics | LBNL CRD; LLNL | Fast computation of volume potentials on structured grids using the method of local corrections. | |
Brian Friesen | 2015 - 2016 | Physics (Astro) | LBNL NERSC | Accelerated the Boxlib framework and representative apps for Cori including addition of Tiling, OpenMP threading, Vectorization and Burst Buffer support. | |
Taylor Barnes | 2015 - 2017 | CalTech | Chemistry | VA Tech (NSF MOLSSI) | Optimization of the NESAP Quantum ESPRESSO application (Hybrid Functional excecution in particular) for the Cori system. Optimizations included OpenMP support, vectorization and adding multiple new levels of parallelism over electron orbitals. |
Andrey Ovsyannikov | 2015 - 2017 | Fluid Mechanics |
Intel | Performance optimization of the Chombo framework and chombo-crunch code for the Cori system at NERSC. | |
Tuomas Koskela | 2016 - 2018 | Physics | University of Helsinki | Accelerated the XGC1 NESAP code for Cori including vectorization support, communications acceleration and the development of the particle in cell (PIC) mini-app ToyPush. | |
Mathieu Lobet | 2016 - 2017 | Plasma Physics |
La Maison de la Simulation (CEA) | Accelerated the Warp NESAP code for Cori including introduction of OpenMP, memory tiling and vectorization for field interpolation, push and charge depositions steps. | |
Tareq Malas | 2016 - 2017 | Computer Science (Stencil Applications) | Intel | Performance optimization of the EMGEO application for the Cori system at NERSC and strategies for MPI communication/computation overlapping. | |
Bill Arndt | 2016 - 2018 | Computer Science (Bio-informatics Algorithms) | LBNL NERSC | Accelerated Cori readiness for HMMER, E3SM, and MPAS codes. Sped up halo neighbor exchanges and threaded memory management in MPAS-Ocean. | |
Zahra Ronaghi | 2017 - 2018 | Biomedical Engineering (Masters in Electrical Engineering) | NVIDIA | Accelerated grid reconstruction methods in the TomoPy tomographic reconstruction code, as well as developed/optimized the Ice Cube machine learning-based event classifier | |
Jonathan Madsen | 2017 - 2019 | Nuclear Engineering | AMD | Accelerated iterative reconstruction methods in the TomoPy iterative tomographic reconstruction codes. | |
Kevin Gott | 2017 - 2019 | Mechanical Engineering | LBNL NERSC | Accelerated Cori and Perlmutter readiness for AMReX and PARSEC. | |
Rahul Gayatri | 2017 - 2018 | Computer Science (HPC Parallelism) | LBNL NERSC | Produced performance portability case-studies performance and models with OpenMP, OpenACC, Kokkos and Raja. Analyzed current state of the art of performance portability. | |
Laurie Stephey | 2017 - 2021 | Plasma Physics / Fusion | LBNL NERSC | Accelerating Python spectroscopic extraction code in the Dark Energy Spectroscopic Instrument (DESI) pipeline | |
Yunsong Wang | 2018 - 2020 | Computer Science (Nuclear Physics Applications) | NVIDIA | ATLAS Cori and Perlmutter readiness including performance analysis of Geant code; multi-node MPI scaling of the ATLAS data analysis pipeline | |
Yan Zhang | 2019 - 2021 | Electrical Engineering (Signal Processing) | Velodyne Lidar | Accelerating progress on NESAP for learning application LSSTNET | |
Muaaz Awan | 2019 - 2020 | Western Michigan University | Computer Science | LBNL NERSC | Accelerating progress on Perlmutter GPU readiness for NESAP application Exabiome |
Brandon Wood | 2019 - 2021 | UC Berkeley | Applied Physics | Developing, tuning, and scaling graph neural networks to accelerate catalyst discovery; part of the Open Catalyst Project. | |
Neil Mehta | 2020 - 2021 | UIUC | Aerospace Engineering | LBNL NERSC | Optimizing particle-based codes (molecular dynamics), C++/Python hybrid codes, Machine learning profiling analysis, Roofline analysis. |
Michael Rowan | 2019 - 2021 | Harvard | Physics | AMD | Improving performance of the Exascale Computing Project advanced particle-in-cell code WarpX, with focus on load balancing. Performance benchmarking and modeling for AI/ML workloads run on HPC platforms. |
Hugo Brunie | 2019 - 2021 | Bordeaux University | Computer Science | CEA (France) | Develop methods and tool to help NERSC users optimize their code with mixed precision tuning optimizations. Current focus on mixed precision tuning for adaptive grid refinement in ASGARD (PI David Green, ORNL). Focus has been made before on NESAP for simulation applications: PeleC (LBL), CCTBX (LBL). |
Dossay Oryspayev | 2019 - 2020 | Iowa State | Computer Engineering | Brookhaven National Lab | GPU optimizations of methods used in the Many Fermion DyNamics (MFDN) application. |
Oisin Creaner | 2019 - 2021 | Institute of Technology, Tallaght, Dublin | Computational Astrophysics | Dublin Institute for Advanced Studies | Enabling GPU simulations of LZ Dark Matter detector. This involves industrial collaboration to upgrade existing Opticks software and fitting this software into the LZ framework. |
Ozgur Cekmer | 2020 - 2021 | University of Tennessee-Knoxville | >Mechanical Engineering | CSIRO | Accelerator performance analysis and optimization of the Adaptive Sparse Grid Discretization (ASGARD) project programming framework. |
Amanda Dufek | 2020 - 2021 | National Laboratory for Scientific Computing (LNCC), Brazil | Computational Modeling | LBNL NERSC | Enhanced roofline modeling including data movement across the HPC interconnect and the GPU bus. |
Raphael Prat | 2020 - 2021 | Bordeaux University | Computer Science | CEA (France) | Stencil-based applications in HPC context. Current project: Optimization of the Proto middleware (used by Chombo4, for example) on GPU supercomputers such as Summit, Tulip or Perlmutter while providing portability and a user-friendly design. |
Jaideep Pathak | 2020 - 2021 | UMD | NVIDIA | Developing novel techniques for augmenting computational fluid dynamics simulations with state-of-the-art machine learning models. | |
Daniel Margala | 2020 - 2021 | UC Irvine | Physics | LBNL NERSC | Porting the DESI spectral extraction pipeline, a Python-based NESAP for Data application, to use GPUs via CuPy and Numba CUDA. |
Muhammad Haseeb | 2023 - 2024 | Florida International University | Computer Science | NVIDIA | C++ based programming model development and evolution. GPU-acceleration of WarpX/AMReX on Perlmutter. |