Postdoc: AI/ML for Uncertainty Quantification

Sandia National Laboratories; California (MO); On-site; USD 85,000 - 140,000; Full time

Benefits offered by this job

Flexible work arrangements

Relocation assistance

Job summary

Sandia National Laboratories is seeking a highly qualified Postdoctoral Appointee to advance AI/ML methods for uncertainty quantification in multiscale materials modeling, and to develop domain decomposition-based hybrid modeling approaches that connect full-order and reduced-order models with data-driven components.

The role requires onsite work and collaboration with principal investigators across national labs and universities, leveraging strong ML, ROM, and computational science expertise.

Qualifications

• PhD in a field of physical sciences, applied mathematics, engineering, or other relevant field conferred within five years prior to employment.
• Knowledge and expertise in machine learning and/or uncertainty quantification.
• Knowledge and expertise in projection-based reduced order modeling and/or operator inference.
• Knowledge and expertise in computational science and/or software development.

Responsibilities

• Develop AI/ML methods for uncertainty quantification in multiscale materials modeling.
• Develop domain decomposition-based hybrid modeling approaches that couple full-order, reduced-order, and data-driven models.
• Research online switching between ROMs and FOMs within subdomains as features evolve, balancing efficiency and accuracy.
• Collaborate with Sandia Principal Investigators and partners at other labs and universities.

Skills

Machine learning

Uncertainty quantification

Projection-based ROM

Computational science

Software development

C++ and Python

Collaboration & communication

Education

PhD in physical sciences / applied math / engineering

Tools

PyTorch

JAX

Julia

C++

Python

Job description

Sandia National Laboratories is seeking a highly qualified Postdoctoral Appointee to advance AI/ML methods for uncertainty quantification in multiscale materials modeling, and to develop domain decomposition-based hybrid modeling approaches that connect full-order and reduced-order models with data-driven components.

The role requires onsite work and collaboration with principal investigators across national labs and universities, leveraging strong ML, ROM, and computational science expertise.