NESAP for Machine Learning Postdoctoral Fellow

Lawrence Berkeley National Laboratory • Berkeley, CA

6d • Hybrid

About The Position

The National Energy Research Scientific Computing Center (NERSC) at Berkeley Lab seeks a highly motivated Postdoctoral Researcher — Scientific Machine Learning (NESAP) to join the Workflow Readiness team as part of NERSC’s Science Acceleration Program (NESAP). You will join a multidisciplinary team building AI-driven scientific workflows for the upcoming Doudna supercomputer. Doudna will deliver over 10x the performance of Perlmutter and connect directly to DOE experimental and observational facilities so research teams can stream and analyze data in near real time. You will collaborate with NERSC staff, domain scientists, and partners at NVIDIA and Dell to prepare high-impact workflows for 12,000+ NERSC users.

Requirements

• PhD in Physics, Chemistry, Computational Science, Data Science, Computer Science, Applied Mathematics, or a related numerical field.
• Programming experience in one or more of: Python, C++, Fortran, Julia.
• Hands-on experience building and training AI models with frameworks such as TensorFlow or PyTorch.
• Ability to succeed in an interdisciplinary team and communicate results clearly in writing and presentations.

Nice To Haves

• Knowledge of GPU architecture and GPU programming.
• Interest or experience in distributed training on large scientific datasets and staying current with new training methods and architectures.
• Experience with performance and profiling tools such as Perftools, NVIDIA Nsight, AMD uProf, or Omniperf.
• Debugging experience with distributed-memory parallel applications.
• Experience with containers (Docker, Podman, Shifter or similar) and modern software practices such as Git, unit testing, CI/CD, and collaborative development.
• Publications in ML-for-Science, HPC, or systems venues (SC/ISC, PPoPP, IPDPS, MLSys, NeurIPS workshops).

Responsibilities

• Contribute to one or more NESAP AI-based scientific workflows targeting NERSC HPC resources, edge resources, and the DOE ESnet network.
• Develop and apply advanced workflow capabilities to improve performance, portability, and productivity of scientific software.
• Collaborate with computational and domain scientists to integrate state-of-the-art AI with simulation and data analysis, including modern agentic approaches.
• Publish and present results in peer-reviewed venues.

What This Job Offers

• Job Type: Full-time
• Education Level: Ph.D. or professional degree
• Number of Employees: 101-250 employees