IHSI affiliates recognized as emerging innovators

In partnership with Entrepreneurship@Illinois, the Office of Technology Management (OTM) introduced the new Emerging Innovator award at Founders Week 2026. This award recognizes faculty members who submitted an innovation report to OTM for the first time during the preceding year.

The award celebrates an important milestone in the innovation journey, recognizing the broader potential of a research discovery and taking the first formal step toward exploring its translation into societal or commercial impact.

Two Interdisciplinary Health Sciences Institute (IHSI) affiliates were recognized in the inaugural cohort of emerging innovators: Yiwen Dong, a professor of industrial and enterprise systems engineering, and Jodi Flaws, a professor of comparative biosciences.

IHSI reached out to Prof. Dong and Prof. Flaws to learn more about their projects and the potential health-related impacts. Prof. Flaws noted that the invention she was part of was driven by a former Ph.D. student, Dr. Jaemin Yang, and his Ph.D. advisor Catherine Blake, a professor of information sciences. Dong and Yang share more about their projects below.

Can you describe the project that you are working to bring to market, and what it took to get to this point?

Dong: Many surgical decisions of patients with movement disorders are often made based on how a patient walks. The traditional process involves in-person visit to a gait lab to take precise measurements, and then a group of experts spend 4-5 hours to translates that data into a detailed report. To make this process more efficient, we developed “MotionDecipher,” an LLM-assisted reporting system that converts precise gait measurements into drafted reports in two minutes. To ensure the system is reliable under this high-stakes scenario, we built a rigorous pipeline that first transforms motion into clinically grounded representations and then uses expert domain knowledge to constrain the way LLM drafts the report. 

To get to this point, I spent a year working at a hospital gait lab as a technician, watching how clinicians measure, analyze, and decide, to design a system that clinicians actually need. Then, we deployed “MotionDecipher" there, and the clinicians provided ratings and feedback without knowing who drafted the report (AI vs. human), across 122 patients, to validate the reliability system. In some cases, AI catches subtle features that the human expert can easily miss. 

Yang: Drug candidates most often fail late, after years of work and enormous investment, because the biological reasons they might fail are scattered across thousands of papers and databases and are rarely connected into one picture. Our project borrows the approach that the nuclear power and aerospace industries use to keep complex systems safe: map every path by which a system can fail, then estimate how likely each path is. We apply that same discipline to biology. Large language models read the scientific literature and pull out cause-and-effect evidence, and an engineering risk model assembles that evidence into a map of the biological failure paths that lead to disease, so researchers can see what is likely to go wrong and why.

The project grew out of my background in nuclear safety engineering and a question of whether those methods could work in biomedicine. With Prof. Catherine Blake's guidance in text mining and Prof. Jodi Flaws's expertise in toxicology, I first built a risk model by hand for how an environmental chemical, bisphenol A, contributes to breast cancer, and tested it against gene expression data from real patients. I then automated the process so it could read tens of thousands of research abstracts and build a far more comprehensive model on its own. This work became my dissertation, and together we filed a patent application through the university's Office of Technology Management. I have since founded a startup, Jungle Biotech, to bring this approach to drug development.

Can you describe the societal impact you hope to ultimately achieve?

Dong: Families often wait months for gait analysis appointments, and there’s a significant shortage of neurologists and movement specialists in the rural areas. "MotionDecipher” mitigates the reporting burden from 4-5 hours to minutes, so clinicians can spend their time on the tasks that matters more, such as sitting down with the family, working out the treatment strategy, deciding what this particular patient needs, instead of letting the paperwork eat it.

Ultimately, we aim to increase the access of gait analysis for all people – individuals can obtain a clinical-grade report anywhere, anytime by simply taking a video of their walking on the phone. To achieve this, we are now developing a patient-facing smartphone app and a clinician-oriented webtool for accessible gait analysis in local care centers and patient homes.

Yang: Most drug candidates that enter human trials never reach patients, and each failure consumes years and resources that could have gone to better options. Our goal is to help drug developers see where a candidate is likely to fail, understand why, and change course before those costs are spent, so that safer and more effective medicines reach patients sooner and at lower cost. The same approach can also help public health researchers understand how environmental chemicals shape disease risk.