7/9/2026 Rudy San Miguel
Ryan Wong, CS PhD, was selected as a 2026 ML and Systems Rising Star for his part as a leading senior student in AI innovation. Wong, who is advised by CS assistant professor Saugata Ghose, was one of only 39 selected by MLCommons from a field of over 170 applicants.
Written by Rudy San Miguel
Ryan Wong, CS PhD, was selected as a 2026 ML and Systems Rising Star for his part as a leading senior student in AI innovation. Wong, who is advised by CS assistant professor Saugata Ghose, was one of only 39 selected by MLCommons from a field of over 170 applicants. The program identifies early-to-late stage and recently graduated PhD students and other researchers to develop community, foster research and career growth, enable collaborations and discuss career opportunities among the rising generation of researchers at intersections of machine learning and systems.
Wong’s projects centered around cooperatively designing new software systems and hardware that can reduce data movement for ML+X applications, where X can be nearly any type of workload. Examples of ML+X include agentic coding tools, molecular and medical discovery tools, and modern video games. His designs improve both the ML and non-ML portions of these programs while still maintaining sufficient flexibility to let us reuse the same hardware for many different use cases, lowering costs significantly.
One of Wong’s projects, “DARTH-PUM: A Hybrid Processing-Using-Memory Architecture,” leverages an emerging approach known as in-memory computing (IMC) to minimize data movement. Recent proposals for IMC hardware often target highly parallel matrix operations that perform repeated multiplication—such as those found in ML workloads. While industrial prototypes have shown that in-memory computing of these multiplications can significantly reduce ML computation time and energy, this hardware is usually a poor fit for non-ML workloads. DARTH-PUM introduces the concept of a hybrid in-memory computing architecture. Wong and his team were able to augment multiplication-capable IMC with the ability to perform any operation that a modern CPU can do, but 2000 times more efficiently.
“With such broad capabilities,” Wong said, “we can now do both ML and non-ML operations extremely efficiently, without the need for most of the data movement to the CPU.”
The other project, “ANVIL: An In-Storage Accelerator for Name-Value Data Stores,” enables computation within a conventional solid-state drive (SSD; the storage disk in modern computers) to quickly look up specific pieces of data of interest from a large pool of information. ANVIL removes the high degree of data movement from the SSD to the CPU that such needle-in-a-haystack lookups used to require by performing the entire process within the SSD. ANVIL makes use of new in-SSD computation techniques to iterate over billions of data pieces per second, significantly bringing down the energy usage and time needed to process both ML and non-ML portions of programs.
Wong said the Rising Star designation completes a circle in his research.
“This recognition feels like it’s acknowledging the circuitous research journey I’ve taken, starting with optimizing data movement for ML-specific workloads, to linear algebra and scientific computing, to graphs, databases, cryptography, and back to [computer hardware] architectures for ML+X,” Wong said.
He also expressed gratitude for the advisors and mentors who have supported him throughout his academic career: Ghose, his advisor; Ben Feinberg and Sapan Agarwal, Sandia National Laboratories, where he was a graduate R&D intern; and Engin Ipek, his research mentor during his time at the University of Rochester.
“Working with this team, and other collaborators, has really broadened my perspective and taught me the value of collaboration,” Wong said. “They always challenge me to think bigger and to consider real-world factors on how different applications will interface and be deployed on our proposed designs.”
As far as doors this commendation will open, Wong says, “I hope this recognition will help me meet other researchers on both the ML and X sides, and open avenues for new collaborations or even new application domains.”
Grainger affiliations:
Saugata Ghose is an Illinois Grainger assistant professor of computer science in the Siebel School of Computing and Data Science.