Graduate R&D Intern: Deep Reinforcement Learning for Multi-scale Coupling

Job Description

Graduate R&D Intern: Deep Reinforcement Learning for Multi-scale Coupling | Sandia National Laboratories

The Tone:
This is an internship at Sandia National Laboratories, located remotely. Sandia is the nation’s premier science and engineering lab, dedicated to national security and technology innovation. This role is crucial for advancing high-level research and development in solid mechanics, structural dynamics, hypersonics, numerical methods, and model order reduction by applying Deep Reinforcement Learning to complex multi-scale coupling challenges.

The TL;DR
• Role: Internship
• Type: Temporary, Full-time
• Location: Remote

• Team: Quantitative Modeling and Analysis Department
• Mission: To conduct research using Deep Reinforcement Learning to develop optimal domain decomposition strategies for hybrid multi-scale coupling.
• Tech Stack: Python

What You’ll Actually Do
• Design and implement a DRL framework for guiding optimal domain decomposition strategies.
• Design and implement a DRL framework for guiding on-the-fly switching between reduced order models and high-fidelity models within a multi-scale domain decomposition framework.
• Perform theoretical analyses of novel coupling and adaptation algorithms.
• Program and test new and existing algorithms and develop and support open-source research software packages.
• Engage with the community for conferences, workshops, proposals, and outreach, and publish research results in high-quality journals and competitive conference venues.

The Must-Haves
• Background: Earned bachelor’s degree and currently attending and enrolled full time in an accredited science, engineering, or math graduate program immediately preceding the internship.
• Experience: Strong background in Deep Reinforcement Learning (DRL), partial differential equation- (PDE-) based modeling and simulation (mod/sim), and adaptive mesh refinement.
• Skills: Minimum cumulative GPA of 3.0/4.0 and ability to work up to 40 hours per week during the summer. Must be legally authorized to work in the United States without sponsorship.
• Bonus: Pursuing a PhD in computer science, mathematics, mechanical engineering, or a related field. Experience with domain decomposition-based coupling methods, Python, non-intrusive Operator Inference (OpInf) ROMs, or the Schwarz alternating method for multi-scale coupling.