Postdoctoral Research Associate - Atomistic Machine Learning

Overview: We are seeking a Postdoctoral Research Associate who will focus on the development, application, and optimization of an active learning workflow for dataset generation of atomic structures and training of machine learning interatomic potentials (MLIPs) in a way that maximizes the capability of high-performance computing (HPC) platforms available in the Oak Ridge Leadership Computing Facility (OLCF). This position resides in the Advanced Computing for Chemistry and Materials Group in the National Center for Computational Sciences (NCCS), Computing and Computational Sciences Directorate (CCSD), at Oak Ridge National Laboratory (ORNL).

Major Duties/Responsibilities:

• Develop, implement, and apply machine learning interatomic potentials (MLIPs) for molecular dynamics simulations.
• Develop and optimize automated workflows for data generation and active learning
  , including molecular dynamics (MD) sampling, configuration selection, labeling with reference methods, and model training.
• Implement and optimize large-scale MD simulations using MLIPs and other atomistic models on HPC resources.
• Evaluate, integrate, and optimize different MLIP architectures and implementations, including pretrained universal models and models tailored to specific chemical and materials systems.
• Develop, debug, profile, and optimize scientific codes and workflows on massively parallel computing systems, including GPU-accelerated platforms.
• Develop and maintain scientific software, primarily in Python
  , using modern practices for software development, testing, documentation, and version control.
• Work with tools and libraries for atomistic simulation and machine learning, such as LAMMPS, ASE, Py Torch , and related tools.
• Develop and evaluate strategies to improve the performance, scalability, efficiency, and robustness of MD simulations, MLIP inference, data generation, and model training.
• Collaborate with computational scientists, HPC specialists, software developers, and researchers across scientific domains to develop and apply new computational methods.
• Conduct collaborative research, analyze and communicate scientific results, and contribute to peer-reviewed journal articles, conference papers and presentations, and technical documentation.
• Deliver ORNL’s mission by aligning behaviors, priorities, and interactions with our core values of Impact, Integrity, Teamwork, Safety, and Service. Promote equal opportunity by fostering a respectful workplace - in how we treat one another, work together, and measure success.

Basic Qualifications:

• A Ph.D. in Chemistry, Physics, Materials Science, Chemical Engineering, Computer Science, Applied Mathematics, or a related discipline
  , completed within the last five years.
• Experience with molecular dynamics (MD) simulations
  .
• Experience with machine learning interatomic potentials (MLIPs) or atomistic machine learning methods applied to chemistry, physics, or materials science.
• Experience with Python and scientific software development.
• Knowledge of fundamental machine learning concepts, including the training, validation, and application of neural network models.
• Experience with Linux-based scientific computing environments for HPC
  .

Preferred Qualifications:

• Significant experience developing, training, fine-tuning, or applying MLIPs
  , including graph neural network-based models, equivariant models, or pretrained universal potentials.
• Experience with MLIP frameworks such as HIP-NN, MACE, NequIP, Allegro, ANI, and UMA
  .
• Experience with atomistic simulation software such as LAMMPS, ASE, GROMACS, OpenMM, Torch Sim
  , or related MD simulation packages.
• Experience with MLIP-based molecular dynamics
  , including system preparation, simulation execution, trajectory analysis, and evaluation of model stability and accuracy.
• Experience with active learning
  , uncertainty quantification, configuration selection, automated dataset generation, or related methodologies.
• Experience developing and managing automated and parallel scientific workflows
  , including task execution systems, dependency management, and pipeline execution on distributed computing resources.
• Experience with HPC systems
  , including running and managing applications on multi-node, GPU-accelerated systems.
• Experience with HPC resource and workload management systems such as Slurm
  , using workflow methodologies and/or tools such as Parallel, Flux or
  Parsl
  .
• Familiarity with concepts of parallelism, scalability, and efficient utilization of computational resources.
• Experience with GPU…