Graduate student Shehryar Khan advances machine learning research

When Shehryar Khan was exploring graduate programs in computer science, Virginia Tech quickly rose to the top of his list.

“I earned my undergraduate degree from Lahore University of Management Science in Pakistan,” Khan said. “Virginia Tech was an obvious choice for my interests, especially applied machine learning. It’s considered a Tier 1 university, and I also had many friends here who spoke highly of the academic culture and community.”

Now a graduate student working toward his Master of Engineering in computer science and applications at the Institute for Advanced Computing in Alexandria, Khan’s research interests span applied machine learning, machine learning optimization, post-training evaluation building, and machine learning security. His work reflects Virginia Tech’s strength at the intersection of advanced computing, research innovation, and real-world impact.

“My work focuses on optimizing and building systems that make research information more organized, accessible, and accurate,” Khan said. “We derive meaningful insights from very messy data sources. I like to say I do ‘research in research.’”

In his second semester, Khan was awarded a graduate assistantship with the University Libraries’ research impact and intelligence team and began working with Assistant Director for Research Intelligence, Engineering Analyst Sarah Over, who oversees the Patent and Trademark Resource Center in University Libraries at Virginia Tech. The role allows him to apply his technical expertise to support research across the university.

“Patents, like other forms of publications, are getting harder and harder for even experts to keep up with in the field, as granted patents per year have more than doubled over the past 15 years,” said Over. “The work Khan proposed and has been working on with me has the potential to tell if a new idea is innovative enough for a new patent.”

Khan is conducting research at the intersection of patents and machine learning under Over's guidance. This emerging area of study explores how machine learning methods can be applied to patent data, opening new possibilities for understanding innovation trends and research impact.

“We are researching how to push models like ChatGPT to produce new ideas. To attempt this, we aim to ask these models to generate patents for us but avoid ideas that already exist,” said Khan. “This is super exciting because present-day models can be naively seen as ‘Oh, they just predict the next word based on what they already know.’ Tackling this in the machine learning space is the next big thing, and we are proposing an idea that hasn't been discussed enough in these spaces. The combination of machine learning, research evaluation, and intellectual property is full of potential.”

“Virginia Tech needs to keep advancing within fields like machine learning and AI [artificial intelligence]. Even if it is just one project, this ties into other work the University Libraries is doing,” said Over. “We can contribute in unique ways with analyses around research publications, patents in this case, which fill a gap in AI-related research done by others at Virginia Tech. To sum up, you might say this is research on research for this type of work.”

Regarding Khan’s work on patents and machine learning, specifically dealing with evaluating novelty in patent submissions, it will likely be published with him as the first author.

“Khan has already contributed to a conference paper that is in peer review and many projects as part of our department, Research Impact and Intelligence,” said Over.

Khan plans to pursue a career in industry, ideally in a machine learning engineering or research-focused role. He credits Virginia Tech — and the opportunity to work with University Libraries — with helping him bridge the gap between advanced theory and applied research.