Dr. Sharlotte Krame of Sandia National Laboratories credit: Randy Montoya
Dr. Sharlotte Kramer is a Distinguished Member of the Technical Staff at Sandia National Laboratories and researcher in Sandia’s Structural Mechanics Laboratory.
She first became involved in HPC-AI in 2013 at Sandia as a member of the Experimental Solid Mechanics Department, where she was a Principal Investigator for a project examining the deformation and failure of metal laser welds.
She received her bachelor’s degree in aerospace engineering from the University of Virginia, and her Master’s and Ph.D. in aeronautics from the California Institute of Technology.
Congratulations to Sandia’s Dr. Sharlotte Kramer, an HPC-AI Vanguard.
Who or what has influenced you the most to help you advance your career path? Is there someone you would like to recognize?
My training prior to working at Sandia focused on development of novel experiments to quantify the mechanical behavior of materials and structures, and I did not have much interaction with my computational counterparts in mechanics.
At Sandia, I learned that the deep integration of experiments and computations provided a richer understanding of mechanics that enable our stakeholders to make critical decisions. My interactions with computational mechanicians changed how I designed my experiments so that the rich data could be used in computational simulations. This then led to me seeking novel ways to incorporate that data using AI methods.
Though I do not run the simulations nor write the AI algorithms, I lead those who do, with the goal of the tight integration of experiments and simulations that ultimately provide a fuller understanding of how these materials and structures behave. I also lead benchmarking challenges (Sandia Fracture Challenge and Sandia Mechanics Challenge) to evaluate the state of the art in computational mechanics, with some teams utilizing HPC resources at Sandia. Dr. William (Bill) Scherzinger and Dr. Edmundo Corona were instrumental in my growth in what experiments and simulations capabilities were needed for HPC simulations of mechanics.
Dr. Kevin Long, Dr. Craig Hamel, and Dr. Dan Bolintineanu opened the door of AI for mechanics to me. Dr. Brad Boyce brought me onto the benchmarking challenge team and gave me the leadership over time. I am truly grateful for their mentorship and collaboration.
What is your passion related to your career path?
My passion is to enable critical decision-making for our national security through high-quality understanding from experiments and simulations. Though I am naturally curious, I do not seek innovation for the sake of innovation, but rather for better service to our nation. As a researcher at a national laboratory, I have the honor of pushing the boundaries of what we can do and learn in mechanical engineering so that others can make the crucial decisions. Our national security is built on strong teaming across disciplines, and I have the pleasure of building such teams that span experimentalists, computational analysts, computer scientists, materials scientists, and data scientists.
Our HPC capabilities enable evaluations of mechanical behavior not possible with experiments alone, and our research in AI has the promise of improving our predictions and enabling optimal experimental design to make best use of experimental and computational resources. I hope my career in such research will be a part of a greater movement to faster and better understanding in mechanics to enable faster and better decisions for our nation.
Sandia National Laboratories
Is there a specific advance related to HPC-AI that you’ve thought about or been exposed to — possibly early R&D work — that shows particular promise by the end of the decade?
Dr. Craig Hamel, in collaboration with Dr. Kevin Long and me, developed a physics-informed neural network (PINN) approach to calibrating material models using full-field experimental surface displacement fields and global loading conditions of a material under heterogenous loading. Rooted in mechanics principles, the PINNs approach does not suffer from the many disadvantages of other inverse methods for material model calibration, such as requiring full volumetric displacement data like the Virtual Fields Method, nor needing slow, computationally intensive iterations like Finite Element Method Updating approaches. This PINNs approach is robust to data drop-outs of the surface data and has the potential to enable calibration of models without an analytical/phenomenological form.
This approach could change how we calibrate material models, reducing the number of required tests and making better use of full-field experimental capabilities. This would in turn improve material modeling in HPC simulations of mechanical behavior of materials and structures.
What are your thoughts on how we, the nation, build a stronger and deeper pipeline of talented and passionate HPC and AI professionals?
We need education on the potential benefits of AI to alleviate routine tasks and support discovery where not previously possible. AI is a tool that will free people to be more creative and focus on the difficult tasks that require human ingenuity. When apprehension of AI subsides, which unfortunately has an unknown time horizon, then there will naturally be a call for people to be proficient at it, creating a stronger and deeper pipeline. In the meantime, we can teach educators on AI so that AI becomes part of everyday curriculum, which widens the pipeline.
We also must teach ethical use of AI so that it is not used maliciously or inappropriately as it becomes more pervasive.
What does it take to be an effective leader in HPC and AI?
As with any tool, curiosity is a key factor in being an effective leader. That leader does not necessarily have to be an expert to appreciate the benefits and foster a work environment, projects, and teams that utilize HPC and AI.
As an experimentalist, I see HPC and AI as a natural part of the workflow that can utilize my data for an end-goal. When first working with computational analysts and data scientists, I entered our collaborations with curiosity on what might be possible and worked with them on the details to bring that to fruition. Another aspect of an effective leader in HPC and AI is to keep the team oriented towards the end-goal and not the HPC and AI tools for their own sake, preventing the team from losing sight on why they were working on the projects. The advancement of HPC and AI is complex, so a clear end-goal is needed to cut through that complexity; an effective leader can rally the team and keep them advancing in the same direction.
Is there an aspect of your work that you are particularly proud of? Have you had a “Eureka” moment that you would like to share with the HPC-AI community?
I am proud of the fact that I am considered an AI professional without having to write any AI code. It may be unusual, but it works for my teams and me. As a data producer, I work with the AI developers to best utilize my data to support our work in mechanics.
Early in my career at Sandia in 2015, I attended an internal workshop on “Big Data,” this is when I realized that my information-rich data with full-field experimental methods would go underutilized unless I championed research for HPC-AI colleagues to better utilize those data. The HPC-AI tools were not going to be tailored to ingest those data unless data producers like me supported their development. This set me on my path to seek such work.
I urge the HPC-AI community to seek out data producers as collaborators at early-stage algorithm development to take best advantage of the data and perhaps even influence how those data are taken, organized, and archived for HPC-AI consumption.
What changes or challenges do you see for the HPC-AI community in the next 5-10 years?
A major challenge for the HPC-AI community is data curation and archiving, whether that be data inputs to the HPC-AI tools or data created by the HPC-AI tools. Data needed to calibrate HPC simulations need pedigree so that the outputs can be confidently used to make critical decisions. Data needed for AI must be curated so that the AI tools can properly use those data and produce reliable inferences. Data produced by HPC-AI need provenance to support decision-making and feed future work beyond their original propose, though that must be done thoughtfully and ethically.
Where we store data, what metadata is stored to ensure the data sets are useable for the application, and how much to store is an enormous challenge when we have limited time and resources to curate and archive. In the next five-10 years, I hope to work with my experimental and computing colleagues to agree on standard approaches for metadata in our technical area, and work to retroactively archive what is deemed the most important to keep for posterity.
Would you like to share any personal information with the HPC-AI community?
Outside my professional pursuits, I have many other interests that show my enthusiasm for learning and personal growth. I have a deep passion for music, I’m trained in classical piano and as a classical mezzo-soprano. I sing in a semi-professional choir in Albuquerque, NM, being a soloist with the NM Philharmonic and at Carnegie Hall. I also lead the worship music at my church.
I travel around the world to learn about different cultures and environments, having been to 14 countries, five continents, and 42 U.S. states. I love to create things with my hands. I sew and make jewelry. I bake often, sharing my creations with my family, friends, and coworkers and sharing my process through pictures online. My passion to learn and share my knowledge can be seen in all aspects of my life.
Sandia New Release Statement:
Sandia National Laboratories is a multimission laboratory operated by National Technology and Engineering Solutions of Sandia LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration. Sandia Labs has major research and development responsibilities in nuclear deterrence, global security, defense, energy technologies and economic competitiveness, with main facilities in Albuquerque, New Mexico, and Livermore, California.
