Postdoctoral Research Associate - AI-Accelerated Discovery of Permanent Magnets

Requisition Id 16541 

 

Overview:

Oak Ridge National Laboratory is the largest US Department of Energy science and energy laboratory, conducting basic and applied research to deliver transformative solutions to compelling problems in energy and security.

 

We are seeking an outstanding Postdoctoral Research Associate with a strong background in condensed-matter physics and materials science – especially related to magnetic materials, experience with first-principles electronic structure methods and proven expertise in developing and/or applying advanced AI/ML methods for accelerated materials discovery. Experience with developing machine-learning surrogates for structure-property relationship, generative AI models, material representations, machine learning force-fields (especially extensions to spinful system) and disordered materials is also desirable. The project will involve developing autonomous materials discovery workflows on HPC platforms that can learn structure-chemistry-property relationship in complex magnets via interpretable machine-learning models, and develop improved AI models that can accelerate prediction of new synthesizable magnet candidates with high energy density and critical temperatures based on these predictions.

 

The position resides in the Nanomaterials Theory Institute (NTI) within the Theory and Computation Section (TACS) at the Center for Nanophase Materials Sciences (CNMS) Division, Physical Sciences Directorate (PSD) at Oak Ridge National Laboratory (ORNL) and will include close interaction with experimental programs at MSTD to synthesize new permanent magnets. The candidate is expected to work closely with Addis Fuhr and P. Ganesh.

 

As part of our research team, you will be working with a highly interdisciplinary team of scientists at the CNMS, MSTD and across other divisions at ORNL.

 

Major Duties/Responsibilities:

• Work closely with members of NTI and CNMS to develop new AI models for discovering novel permanent magnets with targeted properties using advanced concepts such as classifier free guided diffusion models, transformers with multi-headed attention, physics-informed neural networks, materials foundational models with multi-task learning, symbolic regression, reinforcement learning, monte-carlo tree-search, causal ML etc.
• Design, develop, and validate interpretable cross-modal AI/ML models incorporating features from electronic structure theory for predictive structure-chemistry-property discovery in magnetic solids and validate them against multi-modal experimental measurements
• Perform high-throughput first-principles electronic structure calculations (e.g. DFT and post-DFT methods) for generating datasets to train AI models leveraging DOE’s HPC platforms
• Develop new methodologies that can describe both atomic and spin relaxation accurately but at a much cheaper computational cost than DFT
• Present and report research results and publish in peer-reviewed journals in a timely manner
• Ensure compliance with environment, safety, health, and quality program requirements
• Maintain a strong commitment to the implementation and perpetuation of values and ethics
• 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 PhD in Condensed Matter Physics, Materials Science, Chemistry, Physics, or a closely related science discipline completed within the last five years

 

Preferred Qualifications:

• A demonstrated record of developing advanced physics-informed AI models for scientific discovery
• Hands-on expertise developing and applying machine learning for materials and/or process discovery, particularly quantum materials
• Some form of expertise in methods such as machine-learning force-fields for spinful materials, or multi-fidelity Bayesian models that can learn machine-learning force-fields along with effective spin Hamiltonians from ab initio / experimental dataset or machine-learning tight-binding DFT methods
• Expertise in using or developing generative tools for automation of scientific discovery
• Expertise in using high-performance computing (HPC) platforms for delivering breakthrough scientific results
• A