Editor’s note: Shortly after this story was published, Denys Poshyvanyk was named a distinguished member of the Association for Computing Machinery, the world’s largest and most prestigious society of computing professionals. Distinguished members are selected by their peers for outstanding accomplishments that have a significant impact in the field of computer science.
Denys Poshyvanyk, chancellor professor of computer science at William & Mary, has been elevated to Fellow of the Institute of Electrical and Electronic Engineers (IEEE).
“Denys is a prolific scholar who has produced results of impact and consequence,” said Evgenia Smirni, Sidney P. Chockley Professor of Computer Science at W&M. “According to Google Scholar, he is the 10th most cited author from W&M, and his work is of tremendous influence, as shown by his multiple Test of Time Awards for research contributions. He has squarely put W&M on the software engineering map.”
Poshyvanyk is the fourth member of the W&M computer science department to be elevated to IEEE fellow. The other three are Smirni, Qun Li and Gang Zhou.
With approximately 427,000 members in more than 190 countries, the IEEE is the world’s largest professional technical society, consisting of engineers, scientists and allied professionals. Member research focuses on electrical and computer science, engineering and related disciplines. Less than 0.1% of voting IEEE members are selected annually for fellowship.
With Poshyvanyk’s elevation, 18% of W&M computer science faculty members are now IEEE fellows. For perspective, 13% of University of Michigan’s computer science faculty and 12% of Virginia Tech’s computer science faculty hold this distinction.
Poshyvanyk’s fellowship reflects the excellence of W&M’s computer science department and accentuates the vast potential of W&M’s proposed new school which will join the departments and programs of Computer Science, Data Science, Applied Science and Physics. Data is one of the cornerstone initiatives of W&M’s Vision 2026 strategic plan.
“As a department, we are super-excited about the new school,” said Poshyvanyk. “It emphasizes the role of computer science and data science in shaping the future of liberal arts education. The new school will open up even more opportunities for computer science and create meaningful collaboration mechanisms both within and outside of W&M.”
Poshyvanyk’s fellowship is in recognition of his contributions to integrating software analyses and machine learning, a type of artificial intelligence that allows computers to carry out complex tasks without being given precise commands.
Poshyvanyk explained that when developers write source code, it isn’t just the final code that is produced. The changes accumulated throughout the development process are stored in the version tracking systems, including bug reports, evolving documentation requirements and all past versions of the code.
“We can take advantage of all of the artifacts left behind by developers when they’re building the software,” said Poshyvanyk. “We can utilize that information to provide actionable insights for developers so that they can be more productive.”
Poshyvanyk was one of the pioneers in utilizing deep learning and neural large language models (LLMs) for software development. Deep learning is a branch of machine learning that uses data and algorithms to mimic the way that humans learn. LLMs use enormous amounts of data and can generate or review code by recognizing, translating, predicting or generating text.
When Poshyvanyk began this research about 10 years ago, the idea of using deep learning and LLMs for software development seemed far-fetched to many computer scientists.
“My papers were rejected,” Poshyvanyk said. “People were saying, ‘This is science fiction. Do something real.’ My colleagues and I had a deep conviction at the time that this was going to be really useful down the road, and it turns out we were right.”
A 2015 study co-authored by Poshyvanyk showed that deep learning produced higher quality models than those produced by systems that were the status quo at the time. Using deep learning models for code suggestion also produced superior results. In its conclusion, the study proposed future applications for deep learning that are now widely used in software engineering.
Another example of Poshyvanyk’s early influential work is a 2016 study which found that deep learning outperformed the traditional method of detecting code clones, a common problem for software developers. Since then, software developers have adopted deep learning as the most efficient method of detecting code clones, significantly improving overall productivity.
Poshyvanyk’s more recent work has been influential in the research and practice of automated program repair, which involves teaching a computer to modify an existing program in order to fix a bug. Poshyvanyk and his colleagues developed SequenceR. This novel approach to automated repair uses sequence-to-sequence learning, which translates from one language to another, and a copy mechanism to overcome the problem of large vocabulary in source code. This was a huge leap in automated program repair.
“His achievements reflect both the distinction and significance of his work and the excellence of the W&M computer science department as a whole,” said Smirni. “This serves as a harbinger for the success of W&M’s proposed new school and its upcoming graduates.”
Laura Grove, Research Writer