Keynote Speakers

Academic Keynote

Maytal Saar-Tsechansky

Mary John and Ralph Spence Centennial Professor
McCombs School of Business, The University of Texas at Austin

When the Stakes Are High: The AI Advisors We Need and the Research Agenda for Bringing Them into Practice

AI advisors have the potential to transform high-stakes human decisions in critical domains such as healthcare. Yet AI systems used to advise experts do not reliably deliver on this promise. Why can even AI advisors with superhuman expertise fail to reliably improve outcomes in high-stakes contexts, and what would it take for them to add value consistently?

I will first characterize key aspects of high-stakes decision-making contexts that should guide the development of AI advisors. I will explore the outcomes AI advisors produced with these considerations can achieve and whether they exhibit distinct expertise and advising behaviors relative to AI advisors used in practice.

I will then discuss an agenda for developing and deploying reliably value-adding AI advisors across diverse high-stakes contexts. Topics include the risks of inadequate assessment of human and AI decision performance in consequential real-world settings, the curious and consequential case of AI’s selective advising, and the opportunities and challenges of AI persuasiveness.

Biography
Maytal Saar-Tsechansky is the Mary John and Ralph Spence Centennial Professor at the McCombs School of Business at the University of Texas at Austin and a co-founder of Sweetch, an AI-based health company. Her research focuses on advancing AI to improve decision-making and benefit people, organizations, and society. Her recent work focuses on human–AI collaboration, collaboration among AI agents, and trustworthy AI, with the goal of catalyzing AI systems’ positive impact on the world. Beyond AI, Maytal’s research aspires to make a positive impact on problems she cares deeply about, including promoting well-being, mitigating partisan animosity, and promoting democratic attitudes. Her work has been supported by government, industry, and private foundations, and her research has been published in business and science venues, including Science, Journal of Machine Learning Research, Journal of Finance, and Management Science. Maytal has led the University of Texas at Austin’s Translational AI initiative, is an academic board member of the university’s Machine Learning Lab, and is affiliate faculty at The Joe R. and Teresa Lozano Long School of Medicine.

Industry Keynote

Yida Wang, Ph.D.

Senior Director of Compiler Engineering, NVIDIA

Agentic AI for Kernel and Compiler Optimization: From Search to Self-Improving Systems

Modern AI workloads increasingly depend on specialized kernels whose performance is difficult to optimize by hand, especially across rapidly evolving hardware and programming models. This talk examines a new class of agentic methods for kernel and compiler optimization, drawing on three recent directions: self-improving search with optimization memory for accelerator kernels, compiler-agent co-design with structured intermediate representations and localized diagnostics, and reinforcement learning for hardware-aware code generation with robust verification. Together, these works suggest that kernel optimization is moving from manual heuristic tuning toward systems that can reason over program structure, learn from execution feedback, and progressively improve optimization strategies. I will discuss the opportunities this creates for machine learning systems and compiler research, as well as the remaining challenges.

Biography
Yida Wang, Ph.D. is Senior Director of Compiler Engineering at NVIDIA, where he leads a team pioneering compiler and AI agent co-design, exploring how next-generation compilers can evolve to support increasingly intelligent and autonomous AI systems. Prior to NVIDIA, Dr. Wang served as Science Director and Principal Scientist at Amazon's Annapurna Labs, where he led a Machine Learning Systems (MLSys) research team working on AWS Trainium, Amazon's custom AI accelerator. His team focused on optimizing systems and compilers for efficient training and inference of large foundation models, addressing critical challenges in large-scale AI computing. Earlier in his career, Dr. Wang conducted research at Intel Labs' Parallel Computing Lab. He earned his Ph.D. in Computer Science and Neuroscience from Princeton University. With leadership experience spanning NVIDIA, Amazon, and Intel, Dr. Wang brings deep expertise in AI infrastructure, high-performance computing, and emerging agentic AI technologies, offering a unique perspective on the future of AI systems.