Postdoctoral Appointee - Physics-Aware Multimodal Deep Learning

The Advanced Photon Source (APS) (https : / / www.aps.anl.gov / ) at Argonne National Laboratory (Lemont, Illinois, US (near Chicago)) invites applicants for a postdoctoral position to develop physics-aware multi-modal deep learning (DL) methods.

At Argonne, we are developing physics-aware DL models for scientific data analysis, autonomous experiments and instrument tuning.

By incorporating prior physics knowledge into DL model design and training, these models outperform traditional methods even without labeled training data (https : / / www.

nature.com / articles / s41524-022-00803-w). Application spaces for such models include high-resolution 3D imaging, time-resolved materials characterization, and atomic structure determination.

Scientific instrument data is often multimodal in nature and developing DL models that can process and learn from multiple data streams in real-time is key to unlocking the full potential of such instruments.

The postdoctoral appointee will be responsible for developing such methods that are broadly applicable across the physical sciences but applied initially to x-ray characterization needs.

They will publish results in high impact journals, present at conferences and work with the software engineering team to translate the models into production.

The successful candidate will be part of a cross-lab, highly inter-disciplinary team of experts in ML, applied math, HPC, signal processing, computational physics and x-ray science.

The appointee will benefit from access to world-leading experimental and computational resources at Argonne including some of the world’s largest supercomputers (Polaris, Aurora) and one of the brightest synchrotron x-ray sources in the world (APS).

Candidates with a background in deep learning, computational physics, computational materials science, inverse problems, signal processing, x-ray science etc.

are encouraged to apply.

Position Requirements

Required Knowledge, Skills and Experience :

Ph.D. in a related field obtained within the last three years.

Knowledge of x-ray / optical / electron physics, including diffraction, optics, detectors, scattering etc.

Experience with deep learning (DL) libraries such as Tensorflow, PyTorch, JAX etc.

Experience with physics-informed neural networks, automatic differentiation, neural ordinary differential equations, or other physics-aware DL techniques.

Skill in programming languages such as Python, C / C++, Go, Rust etc.

Preferred Knowledge, Skills and Experience :

Experience with version control such as Git and collaborative software development.

Experience with uncertainty quantification and multi-modal deep learning.

Experience with distributed training.

Skill in written and oral communications.

Experience interacting with scientific staff and research groups. Ability to work effectively as a member of a team. Ability to effectively communicate with people of diverse backgrounds and skill sets.

Job Family

Postdoctoral Family

Job Profile

Postdoctoral Appointee

Worker Type

Long-Term (Fixed Term)

Time Type

Full time