Earth Modeling Scientific Programmer

Job Description

Description

The Flexible Modeling System (FMS) is the Earth System Modeling (ESM) framework used at GFDL. The framework supports the multi-physics model components (e.g. atmosphere dynamical core, atmospheric physics, ocean dynamical core and physics, land model, ice model, etc). The FMS core is written primarily in modern Fortran with some elements in C and C++. The core supplies various services that enable grid based domain decomposition, MPI and OpenMP based inter-processor communication, parallel I/O, model diagnostics generation, etc. Importantly, various grids are employed within a single model across the model component space (e.g. “cubed-sphere” for the atmosphere and “tripolar” for the ocean). A recent engagement with the Allen Institute for Artificial Intelligence (AI2) demonstrated the capabilities of utilizing the GridTools for Python (GT4Py) domain specific language to run GFDL’s FV3 dynamical core and associated atmospheric physical parameterizations on GPU-accelerated supercomputers (Pace model). As of the time of publication of this paper, Pace on a GPU performed about 4x faster than the native Fortran version running on CPUs.  The work to date has focused solely on GPU performance and Pace on CPUs runs at about 25% of the natively-coded version in Fortran.

AI2 has transferred this groundbreaking work to NOAA/GFDL and we will be continuing the work in three key areas:

• Integration of the FMS core infrastructure into the GT4Py ecosystem
• Analyze and improve the performance of the GT4Py x86 code generator
• Integrating new backend capabilities, such as Kokkos or MLIR and perhaps LLVM,  into the GT4Py ecosystem

We are looking for an individual with experience in the following areas:

• Python programming for scientific application
• Performance optimization for high-performance computing architectures
• GPU programming
• Knowledge of any of the following is also beneficial:
  • C++
  • Fortran
  • Object-Oriented Python Programming
  • Kokkos Performance Portability Library
  • DaCe - Data-Centric Parallel Programming Framework
  • Domain Specific Languages
  • LLVM Compiler Infrastructure
  • Multi-Level Intermediate Representation (MLIR)

Qualifications

Qualifications:

• A Bachelor’s degree in Computer Science, Information Systems, Engineering, Business or other related scientific or technical discipline
• Two years of experience in Systems Engineering or Programming
• With at least 6 years of specialized experience, a degree is not required
• Significant Experience with Python Programming language
• Experience with GPU programming, HPC architecture, and compilers
• Demonstrated documentation ability in a variety of documentation formats e.g. wiki, google sites, etc.
• Experience with version control tools, such as git
• Support the investigation and debugging of incidents, installation, modification or manipulation of code
• Experience working independently, balancing and prioritizing work
• Ability to obtain and maintain a Public Trust clearance which requires US Citizenship

Overview

SAIC® is a premier Fortune 500® technology integrator driving our nation's technology transformation. Our robust portfolio of offerings across the defense, space, civilian, and intelligence markets includes secure high-end solutions in engineering, digital, artificial intelligence and mission solutions. Using our expertise and understanding of existing and emerging technologies, we integrate the best components from our own portfolio and our partner ecosystem to deliver innovative, effective and efficient solutions that are critical to achieving our customers' missions.

We are approximately 24,000 strong; driven by mission, united by purpose, and inspired by opportunities. SAIC is an Equal Opportunity Employer, fostering a culture of diversity, equity, and inclusion, which is core to our values and important to attract and retain exceptional talent. Headquartered in Reston, Virginia, SAIC has annual revenues of approximately $6.9 billion. For more information, visit saic.com. For ongoing news, please visit our newsroom.