Translating Discovery at Scale: How a Research Software Engineer accelerates science

At Oregon State University, researchers continue to push the boundaries of innovation, driving discovery across science, engineering, art, social science, and humanities. From climate science to robotics to sustainable agriculture, their work is pushing further and delving deeper into some of the world’s most pressing challenges. Advanced technology, like the NVIDIA SuperPod, supports this work, enabling researchers to process, analyze and understand data at unprecedented speed.

As OSU prepares for the Jen-Hsun and Lori Mills Huang Collaborative Innovation Complex and the arrival of the SuperPod, many researchers are already tapping into the university’s high‑performance computing (HPC) resources. However, using these systems requires deep technical skill, typically outside many researchers’ fields. That’s where Research Software Engineers (RSEs) like Chong Chen step in. Chen translates complex scientific ideas into practical software that accelerates research. 

“Many researchers have a deep understanding of scientific problems but need help accelerating their work through proper computing techniques,” he said. “Without HPC, many research projects cannot finish in a reasonable time. My job is to help researchers improve their computing efficiency by utilizing the high-performance computing resources available at OSU.”

Chen has years of experience supporting academic researchers with scientific code optimization and computer efficiency, helping scientists develop applications that solve real-world challenges. He earned his PhD in Electrical and Computer Engineering from the University of Dayton. Prior to joining OSU, he served eight years as an HPC computing specialist at the University of Nevada, Las Vegas. 

While there are other roles at the university that provide similar technical support, Chen is OSU’s first and only RSE. He currently works with researchers in oceanography, forestry, geospatial science, life sciences, and precision agriculture, developing the scientific software that facilitates data collection and analysis. He also assists with grant writing and question development, often gathering the preliminary data that proves the proposed work can be accomplished. Having an RSE on the grant, especially one with a PhD like Chen, can bolster OSU’s competitive edge for grant funding. 

Since joining OSU last August, Chen has supported two major projects, helping scientists work faster without having to navigate the technical complexity of modern computing. In one project, he collaborated with Andrea Jenny, a Climate Sciences Assistant Professor in in the College of Earth, Ocean, and Atmospheric Sciences (CEOAS), and An-Yi Huang, a CEOAS PhD student, to extract storm data from multiple sources. “I accelerated a system developed by An-Yi Huang that pulls weather data from seven different sources and combines them to analyze thousands of storms. Through profiling, we discovered that about 63% of the total time was spent loading and reorganizing data rather than analyzing storms. I redesigned the code structure with parallelism to significantly reduce the data processing time,” he said. “This insight changed how we approached the problem. Scientists know what data they need, but RSEs figure out an efficient path to get it at scale.”

The second project focused on accelerating a climate model developed by Andreas Schmittner, a CEOAS Climate Sciences professor. Climate models, which simulate Earth’s oceans, atmosphere, land, ice, and energy flows using millions of mathematical equations, are computationally intensive and can take months to run, even on supercomputers. Faster models increase the potential for more experiments and exploration, as well as timely decision-making. 

Not only did Chen make the model run quicker, he also troubleshooted a complex technical issue that could have derailed the project. “The tricky part wasn't just making it faster. It was ensuring the faster version gave the exact same scientific answers,” he said. “When I first parallelized it, the code produced wrong numbers because multiple processors were accidentally overwriting each other's data. I had to carefully redesign the code so each processor had its own workspace.”

“Most researchers are not trained on programming languages and computing infrastructure, or trying to leverage computational methods to enhance or help answer questions,” said Chris Sullivan, Director of Research and Academic Computing for CEOAS, who works closely with Chen. “As scientists accomplish work, they may need to go back and optimize the code to be efficient, scale up, or leverage different data. This is where an RSE can step in and quickly identify the areas of processing that need to be adjusted to really take advantage of the hardware and software.”

Chen’s process starts with understanding what a researcher wants to learn. Then he measures where computational time is actually being spent and builds solutions that are correct first and fast second. “The storm data project above is a good example: profiling revealed the real bottleneck was data loading, not the analysis itself.”

As OSU brings the NVIDIA SuperPod online, RSEs will be vital in translating its GPU‑powered capability into scalable solutions for complex research problems. “The SuperPod will give OSU a unique advantage among universities worldwide,” Chen said. “Efficiently utilizing this computing capability is not easy and requires assistance from RSEs like me. I look forward to helping researchers take full advantage of this resource.”

The SuperPod is part of OSU’s centralized research ecosystem, and will serve researchers, students, and faculty across the university’s disciplines and fields of study. A centralized RSE team will make advanced computing more accessible and equitable. “Researchers across campus need computational skills as a service they can pull from when needed and without having to hire within the lab at great cost,” Sullivan said. “Through the new Research Computing Office, RSEs could be shared across projects, allowing for reduced cost and increased research capabilities.”