Hybrid HPC Scientist: Model Optimization for Weather

Lynker; College Park (MD); Hybrid; USD 120,000 - 170,000; Full time

Benefits offered by this job

Healthcare (employee)

PTO

401(k) with company match

ESOP

Tuition assistance

Flexible spending accounts

Employee assistance program

Job summary

Lynker seeks a Scientific Developer for Model Optimization in the DC Metro area with hybrid work. The role focuses on HPC optimization of weather/forecast models using MPAS/UFS, CUDA/OpenMP, and profiling frameworks.

Strong C/C++/Fortran and MPI skills are required; PhD not strictly required but is a plus. Responsibilities include identifying bottlenecks, implementing reproducible, portable code, and guiding colleagues in profiling and GPU acceleration.

Qualifications

• Bachelor's or Master's degree in meteorology, atmospheric science, physical sciences, or related field with modeling experience.
• Experience in HPC, software optimization, and large parallel applications.
• Proficiency in C, C++, and/or Fortran with MPI, and familiarity with GPU programming.

Responsibilities

• Identify and resolve performance bottlenecks in the UFS MPAS-enabled models.
• Develop configurable frameworks for profiling, optimization, and portability.
• Port and test accelerated code across systems ensuring reproducibility.
• Collaborate with teams and mentor colleagues on profiling and GPU development.
• Explore GenAI/LLM tools to streamline software development workflows.

Skills

HPC optimization

C/C++/Fortran

MPI

GPU/heterogeneous programming

Education

Bachelors/Masters in meteorology, atmospheric science, or related field

Tools

Git

CMake

Linux scripting

Profiling tools (gprof/TAU)

Job description

Lynker seeks a Scientific Developer for Model Optimization in the DC Metro area with hybrid work. The role focuses on HPC optimization of weather/forecast models using MPAS/UFS, CUDA/OpenMP, and profiling frameworks.

Strong C/C++/Fortran and MPI skills are required; PhD not strictly required but is a plus. Responsibilities include identifying bottlenecks, implementing reproducible, portable code, and guiding colleagues in profiling and GPU acceleration.