Temple University is a founding member of Pennsylvania's new Keystone AI + Quantum Factory, a statewide consortium designed to expand access to advanced computing for AI and quantum research. One early project, led by Associate Professor Xinghua Mindy Shi, aims to build an agentic AI system that continuously models how dementia progresses over time, giving clinicians a sharper tool for predicting and managing Alzheimer's disease.
The consortium brings together universities, industry partners, and public-sector leaders to address a critical bottleneck: the computing power required for large-scale AI training and simulations. For researchers applying AI to scientific challenges, that infrastructure shift removes a substantial barrier. "Some of the large-scale AI training and simulations that might previously have taken us weeks can now potentially be scaled down to a couple of days or even several hours," Shi said.
An agentic approach to dementia care
Shi's research, "Trustworthy Agentic AI for Personalized Dementia Management," confronts a disease that develops gradually, producing incomplete and shifting signals that make it hard to anticipate a patient's trajectory. Her team is building an agentic AI system that integrates information over time, learns evolving patterns, and explicitly accounts for uncertainty. "What we are building is a system that can continuously integrate information over time and learn evolving patterns, helping to improve disease management and care," Shi said.
The system goes beyond raw predictions by estimating how reliable those predictions are, flagging moments when human oversight is essential. That emphasis on trustworthiness is a central focus. "Clinicians also need to understand how reliable those predictions are, when the system may be uncertain and where human oversight is still needed," Shi explained.
Beyond computing: workforce and economic impact
The Keystone Factory is also built to prepare students and researchers for industries where AI and quantum technologies are becoming foundational. At Temple, researchers working on projects like Shi's now train models on simulations that were previously out of reach due to limited local compute resources. "Initiatives like the Keystone AI + Quantum Factory help expand what we are able to do as computing demands continue to grow," Shi added.
Jen Gilburg, Deputy Secretary for Technology and Entrepreneurship in the state Department of Community and Economic Development, emphasized the role institutions like Temple play in economic competitiveness. "The Keystone AI + Quantum Factory is designed to turn research into new companies, high-quality jobs and economic opportunity across the commonwealth," Gilburg said.
Why this matters for science and research professionals
This consortium model shows how shared infrastructure can shorten the time between hypothesis and result - compressing weeks of model training into hours. The agentic AI system under development does more than predict; it reveals when its predictions are shaky, a feature that becomes mandatory as AI moves into clinical settings. For researchers, the signal is clear: the next generation of scientific computing won't just run faster, it will require architectures that blend scale with trustworthy uncertainty quantification. As quantum computing edges closer to practical use, the human talent pipeline that the Keystone Factory is building will likely define how quickly those tools translate from lab to bedside.
Your membership also unlocks:
