Postdoctoral Researcher – Scientific Machine Learning & Computational Chemistry

Postdoctoral Researcher – Scientific Machine Learning & Computational Chemistry

AM-Applied Mathematics and Computational Research

107376

The Applied Mathematics and Computational Research Division at Lawrence Berkeley National Laboratory (Berkeley Lab) is seeking a Postdoctoral Researcher – Scientific Machine Learning & Computational Chemistry to conduct fundamental and applied research at the intersection of machine learning and computational chemistry. In this role, you will develop physics-informed, symmetry-aware models that accelerate excited-state simulations, integrate them into open-source scientific workflows, and validate their accuracy and performance on energy-relevant molecular and materials systems.

You will also create robust, open-source implementations and connect the resulting models to nonadiabatic molecular dynamics simulation workflows.

We’re here for the same mission, to bring science solutions to the world. Join our team and YOU will play a supporting role in our goal to address global challenges! Have a high level of impact and work for an organization associated with 17 Nobel Prizes!

You will:

Conduct fundamental research in physics-informed and symmetry-aware machine learning for nonadiabatic excited-state molecular dynamics.

Develop and evaluate equivariant graph neural networks and related architectures that learn multiple-state adiabatic potential-energy surfaces, energies, gradients or forces, derivative nonadiabatic couplings, and state-transition behavior from electronic-structure data.

Design reliable data-generation and training workflows, including sampling, active learning, transfer learning, validation, and uncertainty or robustness analysis, to reduce the cost of generating excited-state training data.

Implement, test, document, and maintain research software; integrate trained models with surface-hopping and related nonadiabatic or path-integral workflows and with relevant simulation-code interfaces.

Plan controlled ablation studies and reproducible evaluations using scratch training across multiple seeds, exact train and validation metrics, parity and error analyses, gradient checks, and end-to-end GPU timing; document limitations and extrapolation behavior.

Implement, test, document, and maintain open-source Python/JAX research software; collaborate with researchers to connect trained models to DeepMD, NEXMD, i-PI, MixPI, or related nonadiabatic and path-integral workflows and validate.

Publish results in peer-reviewed journals, present at scientific conferences and project meetings, contribute to software releases and reports, and mentor or collaborate with students and researchers as appropriate.

We are looking for:

PhD degree, within the last 3 years, in Computer Science, Computational Science, Chemistry, Physics, Applied Mathematics, Materials Science, Chemical Engineering, or a related technical field.

Demonstrated research experience in machine learning or deep learning for scientific or atomistic data, computational chemistry, computational physics, or a closely related area, with the ability to work across disciplinary boundaries.

Strong Python programming skills and experience developing, training, and evaluating neural-network models with a modern framework such as PyTorch or JAX.

Knowledge of graph neural networks, geometric deep learning, invariant or equivariant models, or related approaches for learning from molecular or physical systems.

Ability to develop reliable research software in a Linux environment using version control, testing, documentation, and reproducible computational workflows.

Record of scientific publication or presentation; ability to conduct independent research, work effectively in a highly collaborative multi-institutional…