Latest IBM Spyre Accelerators Power New UAlbany AI Research Projects

A photo of a black server cabinet with four silver racks illuminated blue from within.
The new IBM Spyre Accelerators installed last week in UAlbany's Data Center will power the next round of research projects funded by the Center for Emerging AI Systems. (Photo by Patrick Dodson)

ALBANY, N.Y. (Jan. 12, 2025) – The University at Albany today announced seven new joint research projects that will use the latest IBM artificial intelligence hardware installed at UAlbany.

The projects are funded through the UAlbany-IBM Center for Emerging Artificial Intelligence Systems and as part of the collaboration between the university and IBM use a cluster of IBM Spyre Accelerator cards for projects ranging from the search for mutational signatures important to diagnosing cancer to the impact of methane emissions on Earth’s climate.

The IBM Spyre Accelerators installed at CEAIS are updated versions of the prototype IBM AIU Spyre conceived in the IBM AI Hardware Center and installed at UAlbany in early 2024 — the first such installation anywhere other than IBM’s own facilities. Built for efficient, low-latency AI inferencing, the Spyre Accelerator is a commercial system-on-a-chip developed for IBM z17, Power11 and LinuxONE 5 systems

A close up of the circuitry and components inside a computer server rack.
The new IBM Spyre Accelerators are the latest versions of prototype IBM hardware first installed at UAlbany in early 2024. (Photo by Patrick Dodson)

“UAlbany’s rich research ecosystem is the perfect testbed for IBM’s cutting-edge AI hardware, allowing researchers at both institutions to showcase what these impressive new systems are capable of while pursuing science that benefits society,” said Thenkurussi “Kesh” Kesavadas, UAlbany’s vice president for research and economic development. “These kinds of industry-academic partnerships are essential for innovation and are central to promise of UAlbany’s AI Plus Initiative to put the latest technology in the hands of our students, faculty and staff. They also provide our faculty exposure to industry-driven research, which is increasingly a major focus for UAlbany.”

"Our collaboration through the Center for Emerging Artificial Intelligence Systems provided valuable learning to progress the field of AI,” said Mukesh Khare, GM of IBM Semiconductors and VP of Hybrid Cloud Research, IBM. “Now, with the next generation of IBM Spyre Accelerators installed at UAlbany, it’s exciting to see how researchers will leverage AI to advance research in medicine, climate science and so many other important areas.” 

Understanding Climate, Cancer and More Efficient AI

The seven newly funded projects are:

  • Deep Learning to Coexist: Spyre-Accelerated Multi-Task Wireless Waveforms: Associate Professor Hany Elgala, Department of Electrical & Computer Engineering — Using IBM’s Spyre Accelerator to learn new wireless transmission signals that can perform multiple tasks at once, such as sending data to a drone while also detecting its presence.
  • Discovering Algebraic Structures towards Automatic Symmetry-aware AI Models: Assistant Professor Halyun Jeong, Department of Mathematics & Statistics — Building, designing and implementing efficient machine learning algorithms that run on the IBM Spyre Accelerator to reliably learn hidden patterns and symmetries in data. Through this process, we can make AI faster, more efficient and more trustworthy.
  • Accelerating Mutational Signature Extraction Using IBM Spyre Accelerators: Assistant Professor S M Ashiqul Islam, Department of Epidemiology & Biostatistics — Developing faster and more efficient methods for analyzing cancer mutations. By speeding up how scientists identify mutation patterns, this project aims to improve understanding of how cancers develop and would support the design of more targeted treatments.
  • Scalable Watermarking for LLM Outputs: IP Protection and Attribution on IBM Spyre Accelerators: Assistant Professor Phung Lai, Department of Cybersecurity — Developing invisible markers, known as watermarking, for AI-generated text to enable reliable source verification, enhancing safety, trust and protection against misuse and intellectual property infringement.
  • Accelerating Post-Training Reasoning and Optimization with Neural Network Approximation: Assistant Professor Chong Liu, Department of Computer Science — Speeding up the reasoning and decision-making of AI systems, such as large language models, by training a lightweight neural network to approximate complex computations. With this shortcut learner, AI can reason and make smart decisions much faster while using less energy.
  • Scalable and Expressive Attention Mechanisms for NLP: Associate Professor Penghang Yin, Department of Mathematics & Statistics — Developing new AI algorithms that enable language models to process much longer text efficiently while using less computing power. This project focuses on improving the attention mechanism, the core part of an AI system that determines which pieces of information to focus on, making it faster and more energy efficient.
  • Quantify Methane Emissions in Foundation Model and Compact-Methane AI Models: Research Faculty Member Xueying Yu, Atmospheric Sciences Research Center — Developing a novel AI framework to advance the accuracy of methane emission monitoring and integrating observations from multiple satellite sensors.

UAlbany and IBM jointly formed the $20 million Center for Emerging AI Systems in 2023 to create a testbed for projects using the latest IBM hardware designed to support AI computing with greater speed and energy efficiency through lower-precision calculations.

Each joint project teams a UAlbany and IBM researcher together. The first five projects launched in 2024.