Data resilience innovators named Association for Computing Machinery Distinguished Members

Computer scientists Bogdan Nicolae and Sheng Di of the U.S. Department of Energy’s (DOE) Argonne National Laboratory have been recognized as Distinguished Members of the Association for Computing Machinery (ACM) for significant and lasting contributions to the computing field.

This year’s prestigious award recognized 61 honorees chosen for their exceptional accomplishments, representing the wide range of research and practice across computing and information technology.

Nicolae was honored for contributions to data management, storage and resilience of high performance computing (HPC). Di was recognized for contributions to lossy compression, HPC resilience and distributed computing.

The ACM Distinguished Member program recognizes up to 10% of ACM’s worldwide membership and requires at least 15 years of professional experience.

Bogdan Nicolae

With innovations that are reshaping how the world’s fastest supercomputers manage and move data, Nicolae is leading a new generation of research at the intersection of HPC and AI.

Nicolae, who joined Argonne in 2017, is a computer scientist in the Mathematics and Computer Science (MCS) division and a research professor at the Illinois Institute of Technology in Chicago. He is also a scientist-at-large at the University of Chicago within the Consortium for Advanced Science and Engineering.

His work is driven by the rapidly increasing complexity of AI systems, where exascale platforms such as Argonne’s Aurora supercomputer rely on tens of thousands of graphics processing units to train a single model. When data cannot move quickly enough through these systems, bottlenecks can cause processors to idle, degrade performance and limit the efficiency of large‑scale scientific workloads.

“As traditional data management breaks down for AI applications, it becomes critical to address the need for innovative frameworks capable of coordinating efficient I/O access at scale, providing advanced capabilities specific to AI patterns and handling resilience,” said Nicolae. ​“Argonne offers a unique front-row seat to these challenges. Argonne supercomputers are not just productivity tools but are essential laboratories where theoretical data and resilience models are stress-tested against the world’s most demanding data-intensive workloads.”

In 2021, Nicolae introduced DataStates, a data model designed to efficiently store and process the massive datasets produced by supercomputers and scientific instruments.

DataStates expands traditional approaches by defining reusable intermediate data states that can be tagged with properties and constraints, enabling the runtime to automatically optimize data placement and movement across complex, heterogeneous storage systems.

This effort earned Nicolae the DOE Early Career Research Program (ECRP) award and has since become a foundation for advances in large‑scale data management.

Building on that foundation, Nicolae leads the Very Large‑Scale Checkpointing and Capture project, which introduces a breakthrough approach for capturing a system’s internal state — the data, progress and configuration needed to correctly resume computation — using a set of composable state providers.

These innovations are already shaping next‑generation AI workflows.

DataStates‑LLM extends the broader DataStates framework with capabilities tailored to accelerate and streamline large language model (LLM) and transformer training.

“In short, my work provides the data backbone required to train and run inferences with massive LLMs and transformers that can hypothesize and predict scientific outcomes at speeds previously thought impossible,” Nicolae said.

Nicolae has served as a mentor to dozens of doctoral students, postdoctoral researchers and computer science interns.

Nicolae, who earned his doctorate in computer science at the Université de Rennes, France, in 2011, has published more than 120 papers in leading scientific conferences and journals. He is an associate editor of the Institute of Electrical and Electronics Engineers (IEEE) Transactions on Parallel and Distributed Systems journal.

Nicolae serves on program committees and helps organize leading international conferences in parallel and distributed systems. This includes the IEEE International Parallel and Distributed Processing Symposium and the ACM International Symposium on High‑Performance Parallel and Distributed Computing. He is a Senior Member of IEEE.

Sheng Di

Since joining Argonne in 2013, Di has built a reputation as one of Argonne’s most influential contributors to large‑scale computing. He has served as a computer scientist in MCS since 2019.

His research spans lossy compression, fault tolerance, scalable HPC and distributed systems. Di’s algorithms have been central to developing world‑class exascale platforms, including Argonne’s Aurora and DOE’s Oak Ridge National Laboratory’s Frontier.

A core focus of Di’s research is his pioneering work in error‑bounded lossy compression, which shrinks massive scientific datasets using small, precisely controlled errors. This approach makes data movement, storage and analysis feasible at scales that would otherwise be out of reach.

Di is the co‑founder and technical lead of the widely adopted SZ Lossy Compression framework which addresses the growing challenge of simulations and experiments that generate data far faster than storage systems can manage.

“The SZ framework gives users explicit, fine‑grained control over numerical error, allowing them to balance accuracy and data reduction in a transparent and trustworthy way,” Di said. ​“By offering predictable error bounds and preserving essential scientific features, SZ helped shift lossy compression from a perceived risk to a reliable, controllable technology for large‑scale science.”

Introduced in 2019, the SZ software is downloaded more than 1,000 times each year by users in the U.S. and across the globe.

Di has made substantial contributions to resilience — the ability of computer systems to continue functioning and recover quickly when disruptions occur.

In 2014, Di helped design a data-analytics-based approach for detecting silent data corruption, representing one of the earliest statistical methods for identifying subtle computational errors that escape hardware-level checks. Over the past decade, this work has been widely recognized and extensively cited across the research community.

He was also among the first to apply transformer‑decoder models to predict system errors, creating one of the field’s most effective machine learning-based resilience approaches.

During his career at Argonne, Di has built a reputation as an influential mentor, guiding many students and early‑career researchers who have since moved into tenure track and tenured faculty roles.

“I’m grateful to work with so many talented students, and I always respect their ideas, give them the freedom to explore their own directions and support them with patience and training,” Di said. ​“Creating a positive, enjoyable research environment helps my students grow. Many have gone on to faculty positions and continue collaborating with our group.”

A leader in professional service, Di has served on program committees for more than 40 IEEE and ACM events and as a reviewer for more than 100 leading conferences and journals. Di has published more than 200 peer-reviewed papers since 2007.

Among his many honors, Di received two R&D 100 Awards: one in 2019 for his contributions to resilience research and another in 2021 for his innovation in SZ compression software. He also is the recipient of a 2021 DOE ECRP award supporting his research on scalable scientific data reduction for scientific datasets. Over the past five years, this project has resulted in more than 90 peer-reviewed publications.