UGA Workshop Charts Path Toward AI-Ready Spectroscopy Data Infrastructure

The University of Georgia hosted a one-day workshop, “Spectroscopy Data and AI: Toward Integrated Research Infrastructure,” on August 21, 2026, bringing together leading researchers in spectroscopy, artificial intelligence, machine learning, data systems, and research cyberinfrastructure to discuss how the scientific community can build shared, trustworthy, and AI-ready spectroscopy data resources.

Held at the UGA Center for Continuing Education, the workshop drew 76 participants from 15 universities and national laboratories across the United States. Participating institutions included Rice University, the University of Illinois Urbana-Champaign, the University of Chicago, Georgia Tech, Emory University, Virginia Tech, the University of Notre Dame, Brookhaven National Laboratory, the National Institute of Standards and Technology, and the University of Georgia. The workshop addressed a growing challenge in spectroscopy: spectral data are being generated at increasing speed and scale, but much of this information remains difficult to reuse because it is stored in proprietary formats, processed through nonstandard workflows, separated from essential metadata, or published only as figures. Participants discussed how open-source platforms, standardized metadata, reproducible workflows, quality-control methods, secure data-sharing models, and AI-enabled tools could help transform spectroscopy from isolated measurements into a more integrated scientific discovery ecosystem.

Prof. Yiping Zhao, Distinguished Research Professor in the Department of Physics and Astronomy at UGA, opened the workshop by outlining the need for a shared spectroscopy data infrastructure based on the FAIR principles: findable, accessible, interoperable, and reusable. He introduced SpectraGuru, an open-source spectral analysis and data platform being developed by his group, as one prototype for supporting reusable Raman and SERS data, standardized workflows, machine-learning analysis, and community contribution.

Dr. Ron R. Walcott, UGA’s interim senior vice president for academic affairs and provost, delivered welcome remarks, emphasizing the importance of bringing together spectroscopy, artificial intelligence, data science, and research infrastructure communities to address emerging scientific challenges.

The keynote lecture was delivered by Prof. Naomi J. Halas of Rice University, a member of both the National Academy of Sciences and the National Academy of Engineering. Her talk, “Machine Learning-Enhanced Surface-Enhanced Spectroscopies for Detecting Environmental Toxins,” highlighted how physics-informed machine learning can be combined with vibrational spectroscopy to detect environmental toxins, including polycyclic aromatic hydrocarbons, PFAS, and microplastics in complex biological and environmental samples.

Invited presentations throughout the day covered a broad range of topics at the intersection of spectroscopy and AI. Prof. Rohit Bhargava of the University of Illinois Urbana-Champaign discussed integrated spectroscopic and AI pipelines for chemical histopathology. Dr. Ozgur Ozan Kilic of Brookhaven National Laboratory presented AI-based scientific workflows for accelerating experiment-to-decision cycles. Prof. Taeho Jung of the University of Notre Dame discussed zero-trust data-sharing architecture for research data management. Dr. Young Jong Lee of NIST highlighted the importance of measurement quality, standards, and reference data for coherent Raman and infrared microscopy. Prof. Haonan Lin of Georgia Tech and Emory University presented advances in computational chemical imaging. Dr. Xiaohui Qu of Brookhaven National Laboratory discussed interpretable machine-learning approaches for X-ray absorption spectroscopy. Prof. Peter Vikesland of Virginia Tech presented AI-assisted SERS analysis for environmental and biological detection. Prof. Anindita Basu of the University of Chicago joined remotely to discuss Raman spectroscopy and metabolite analysis for inflammatory bowel disease and mucosal healing.

The workshop also included a poster session featuring 22 posters, giving students and postdoctoral researchers the opportunity to present their work and interact directly with invited speakers and participants.

The day concluded with a panel discussion focused on whether the spectroscopy community needs a shared AI–spectroscopy infrastructure, what core components such an infrastructure should include, how spectroscopy data systems can evolve into AI-enabled discovery platforms, how spectroscopy should connect with other characterization technologies, and who should support, govern, and sustain such a system.

Several themes emerged from the workshop discussions. Participants emphasized that measurement quality sets the ceiling for any AI model, that reproducible workflows are as important as algorithms, and that scalable data sharing requires clear governance, contributor credit, access control, and community standards. SpectraGuru and related platforms were discussed as possible prototypes for broader AI-ready spectroscopy infrastructure.

The workshop also identified several possible next steps, including a multi-authored white paper or perspective article, a metadata working group, benchmark datasets, shared database pilots, multi-institutional funding proposals, and a recurring workshop series.

The event was supported by the National Science Foundation Pathways to Open-Source Ecosystems (POSE) Program under Award No. 2518273.