Talk titles and abstracts to be announced. Check back for details as they are confirmed.
--

--
https://ncsu.zoom.us/j/98389016656

Ashley Kurian: Protecting Neural Network Intellectual Property: Attacks and Defenses Against Model Stealing

Model stealing attacks on AI/ML devices undermine intellectual property rights, compromise the competitive advantage of the original model developers, and potentially expose sensitive data embedded in the model’s behavior to unauthorized parties. Given the significant efforts of collecting datasets and training neural networks on ML accelerators or other hardware platforms such as FPGAs, GPUs, and non-commercial custom ASICs, the confidentiality and privacy of the trained models need safeguarding. The model stealing attacks can be categorized into two classes: (i) hyperparameter stealing attacks where the adversary aims to learn the architecture of the trained models such as the types of layers and their configurations, and (ii) parameter stealing attacks where the adversary aims to learn the trained weight and bias values. We analyze the model stealing attack landscape and build defenses that can be deployed to mitigate these threats while preserving model utility.

Ahmad Ali: DAEs Are Not ODEs: A Geometric View of Solver Convergence in Switched Dynamical Systems

Differential-algebraic equations (DAEs) arise naturally in the modeling and simulation of many physical systems. However, their numerical behavior can differ fundamentally from that of ordinary differential equations (ODEs), especially in systems subject to discrete switching events. This talk presents a geometric view of convergence in switched DAEs, showing how a change in algebraic constraints redefines the solution manifold, leaving the pre-event state off the new manifold and causing conventional Newton-based reinitialization to fail. To address this, a Homotopy-continuation-based re-initialization strategy is presented that constructs a numerical bridge between the pre-event and post-event solution manifolds. These ideas are demonstrated using power-system-based examples involving converter control-mode switching and network topology changes.
--
Ashley Kurian is a doctoral student working with Aydin Aysu at the Hardware Embedded Cyberthreat Research (HECTOR) Lab. Her research operates at the intersection of hardware security and machine learning, where she uncovers emerging vulnerabilities and builds defenses. She demonstrated how to extract model secrets from black-box commercial accelerators such as the Google Edge TPU, and engineered training-time defenses that counter these threats with zero inference-time overhead. Through her research, she aims to help ML keep its secrets.

Ahmad Ali is a Ph.D. student in the Department of Electrical and Computer Engineering at NC State University, where he is advised by Hantao Cui. He received his M.S. in electrical engineering from Oklahoma State University. His research interests include power system modeling and simulation, applied numerical methods, solver design, and computational acceleration. Most recently, he interned at Google X, where he worked with a computation team developing high-performance simulation tools.
--
https://ncsu.zoom.us/j/98389016656

Quantum technology holds many revolutionary promises, such as exponentially speeding up intractable computation tasks, secure quantum networking, and surpassing the standard quantum limit for sensing and spectroscopy. Scaling up the quantum system is one of the critical challenges in many of these promises. In this talk, I will introduce our recent efforts to address the scalability challenge through photonic integration and optical multiplexing. In our experiments, we generated quantum optical frequency combs in an ultra-high Q optical microresonator on a photonic chip. The optical microresonator can provide thousands of frequency multiplexed quantum modes, where unconditionally entanglement can be created through the Kerr parametric process. I will discuss our ongoing efforts of multipartite entanglement generation, heterogeneous integration of multiplexed quantum source and detection, and their future applications in quantum computing, quantum networking, and quantum sensing.
--
Xu Yi is an associate professor in the Department of Electrical and Computer Engineering and courtesy associate professor of physics at the University of Virginia. He obtained his Ph.D. in applied physics from the California Institute of Technology in 2017. His research interest has been focused on integrated photonics, quantum optics, optical microresonators, optical frequency combs, and their applications in quantum computing and microwave and mmWave photonics. Yi is the recipient of a NASA group achievement award in 2017, an Air Force Young Investigator Program award in 2021 and an NSF CAREER Award in 2022.
--
https://ncsu.zoom.us/j/98389016656

Quantum technologies are advancing rapidly, offering the Department of Energy community a pathway to scientific discovery and quantum utility by 2030. Harnessing quantum technologies as they scale up will require next-generation software and integration with HPC and AI. Software frameworks to couple HPC and AI to quantum computers, as well as connecting end users to such a hybrid computing environment will be essential to ensure the technology can be harnessed optimally to deliver on industry and scientific discovery needs. I will show research progress and discuss the opportunities for scientific discovery with quantum technologies and potential pathways to integrate HPC, AI and networking.
--
Bert de Jong is the director of the Quantum Systems Accelerator, part of the National Quantum Initiative. He is also team director of the Accelerated Research for Quantum Computing Team MACH-Q, funded by DOE ASCR and focused on developing software stacks for near-term quantum computing devices. He has a program in AI and machine learning to understand biomolecular processes and discover new materials and molecular crystals for gas adsorption. De Jong serves as department head for computational sciences and leads the Applied Computing for Scientific Discovery Group, which advances scientific computing by developing and enhancing applications in key disciplines and developing HPC, quantum and AI tools and libraries.
--
https://ncsu.zoom.us/j/98389016656

Engineering education often focuses on solving well-defined problems with known constraints and measurable answers. Real-world engineering is frequently less orderly: information is incomplete, environments change, models are imperfect, and decisions must be made before every uncertainty can be resolved. This talk explores how engineers can approach complex, adaptive problems by combining technical depth with systems thinking, experimentation, and the ability to reason under uncertainty. Using an autonomous robot as an accessible example, we will examine how seemingly straightforward engineering problems become challenging when sensing, resources, objectives, and the environment interact—and what skills engineers can develop to succeed in these environments.
--

--
https://ncsu.zoom.us/j/98389016656