Erin Acquesta of Sandia
In our ongoing Vanguards of HPC-AI series, we now feature Erin Acquesta, who holds a PhD in mathematics from North Carolina State University. She became involved in HPC-AI in 2014, when she worked as a data scientist at Lenovo. In 2016, she joined the team at Sandia National Laboratories as a senior member of the technical staff.
She is highly regarded by her peers and management in her role as a principal research scientist and team leader. She refers to herself as an extroverted mathematician and is passionate about being an effective team leader who values trust and patience as core leadership values.
We are thrilled to congratulate Sandia’s Erin Acquesta on being recognized as an HPC-AI Vanguard.
HPC-AI Current and Future Leaders
Erin C.S. Acquesta
Principal Research Scientist
Sandia National Laboratories
What is your passion related to your career path?
The technological advancements of HPC systems, randomized algorithms, and artificial intelligence have provided novel computational capabilities that are revolutionizing our understanding of complex systems in support of decision science under uncertainty. My career path and passion are focused on establishing trustworthy evidence towards the responsible and reliable use of AI for high consequence applications. As a mathematician, I believe this foundation will be established through a better understanding of the theory that underpins our capabilities; so that we can answer the question “why is this working” instead of simply “how it works.”
Do you prefer working as an individual contributor or a team leader?
I prefer a healthy balance of both. As an extroverted mathematician I have a passion for developing collaborative, multidisciplinary teams. As a team leader, I have the opportunity to set the culture and the expectations on our team that takes in all perspectives that need to be part of the conversation. When I lead these teams, I see the greatest opportunity for innovation. Alternatively, I find that if I focus all my time on developing others and interdisciplinary teams it is hard to keep up with the most recent methods and technologies. Having an opportunity to contribute to a peer project as an individual contributor provides me with the time to continue to grow and keep up with the fast-paced field of AI.
Who or what has influenced you the most to help you advance your career path in this advanced computing community?
I have been so blessed throughout my career to engage with several experts that all have been extremely generous with their time. Even then, the greatest influence I have ever received came from my high school teachers and undergraduate professors. I would not be here today and part of the exciting culture of revolutionary science at the National Laboratories if it wasn’t for the encouragement I received early on.
Math didn’t challenge me in high school. It was my high school teacher, Mr. O’Leary, who took the time to encourage me to pursue a degree in mathematics. He saw the potential in me that I couldn’t see myself. I didn’t take his advice initially, but when I found myself re-evaluating my undergraduate major his words stuck with me and I chose to get a degree in mathematics.
I am a first-generation college graduate. My bachelor’s degree was more than enough. I had no intention of going on to graduate school. Until I went to Ithaca College and Dr. Maceli and Dr. Brown both encouraged me to pursue a Ph.D. in mathematics.
What are your thoughts on how we, the nation, build a stronger and deeper pipeline of talented and passionate HPC and AI professionals?
The passion for AI is already there, we need to balance it with the understanding of the hardware and infrastructure it takes to implement AI responsibility on HPC systems. We need more programs introducing HPC earlier in students’ academic careers.
What does it take to be an effective leader in HPC and AI?
To be an effective leader in HPC and AI, I focus on emotional intelligence with an emphasis on patience.
Emotional Intelligence: To lead innovation for AI on HPC systems will inevitably require a multidisciplinary team. Finding an effective strategy that allows a highly interdisciplinary team of experts from a diverse set of backgrounds and a variety of personalities to collaborate requires a significant amount of emotional intelligence. The approach that I take is rooted in scientific improv with the following 4 key principles:
- I believe that the whole is greater than the sum of the parts and that the secret to success is building the whole.
- I start with building trust with each individual team member. This takes time, but in the end – as the leader – you need to understand each team members strengths, priorities, and goals. This way I can effectively champion their role on the team and hopefully earn their trust in return.
- Then, once each part of the team has established the core components that will lead to their success, we host an in-person (hybrid if necessary) meeting to bring the parts together.
- If trust was successfully built in step 2, then you can expect everyone to show up. Through that trust, I facilitate disagreements with scientific improv. It’s a misconception that you must “agree” with whatever is said in improv. While that is true, you must be careful about what it means to “agree”. In scientific improv, I never ask anyone to blindly agree with someone else’s statement. Instead, I ask that they agree that the statement is something that the person who shared it believes and if it does not align with your own understanding to be curious and want to learn more about why there seems to be a disconnect there. Ultimately, we need to keep the conversation going.
Patience: This is not just patience with people, but patience with finding the right solution to a challenge you come across implementing AI on HPC systems. There are so many things that can go wrong when you implement randomized algorithms on GPUs at exascale. It is not enough to just be smart, or a good problem solver, or even patient. You must have the patience to identify the problems that need to be prioritized. Often, it can be easy to find the “band-aid” solution and move on, it takes patience to dig deeper and get at the root of the problem.
What is the biggest challenge you face in your current role?
Keeping up with the fast-paced development of AI. Open-source libraries and access to GPUs on our laptops is generating new research directions faster than we can evaluate the credibility of the current methods.
What changes do you see for the HPC / AI community in the next five-10 years, and how do you see your own skills evolving during this time frame?
I expect that there will be a new field of mathematics developed that will focus on the theoretical foundation of AI. While some theories exist, e.g., the Universal Approximation Theory, we lack a more formal rigorous foundation that transcends theory in a way that provides more responsible use of AI.
What is your view on the convergence and co-dependence of HPC and AI?
Advancements in HPC are providing opportunities to build larger and more complex models with more and more hyperparameters. As this continues, we keep moving away from our foundational understanding of why the model is performing well or why it does not perform as expected. As we add more and more parameters and hyper-parameters to a model we continue to make the verification, validation, uncertainty quantification, and sensitivity analysis of that model significantly more difficult. Every time we add another component, feature, or capability to a model, you will inevitably break a prior assumption.
Do you believe science drives technology or technology drives science?
I believe the pursuit of objective facts (knowledge) drives both technology and science.
During the scientific revolution in the 16th-17th century, science drove technology. The technology was needed for scientists to standardize their data collection procedures to validate their theories. Today, we see that technology is driving the science. The advancements of HPC and AI resources provide novel exploration into vast amounts of data that is producing new discoveries and requiring a more formal scientific understanding. As technology advances, we have to be able to benchmark the performance of these tools against well-established scientific principles so that we can collect verification evidence that the technology is performing as expected.
In the pursuit of objective facts, technology and science are coupled in such a way that we evaluate the credibility of one against the other. Independently, both technology and science are full of uncertainties and errors. Coupling science and technology is the key to establishing irrefutable knowledge that is extensively verified and validated.
Would you like to share anything about your personal life?
Educational Outreach: VEX VRC Robotics program in New Mexico. Throughout the school year, I am a mentor to the two Cibola High School robotics teams. As a small community that is actively growing, I also supported the broader program by refereeing the majority of the events in New Mexico and as the Head Referee for the Albuquerque League Championships and States Championships.
Hobbies: Reading, traveling, baking, cooking and camping.
Find your next career at Sandia Labs: sandia.gov/careers
