Bridging AI and Healthcare: Dr. Carl Yang’s Mission to Personalize Patient Care

When Dr. Carl Yang joined Emory University’s Department of Computer Science in 2020, he brought with him a deep expertise in data mining and artificial intelligence. Just a year later, he joined the Emory Global Diabetes Research Center (EGDRC), where his interest in applying AI to real-world health challenges began to take shape.

“I always wanted my work in data mining to be more than just technical,” Yang said. “I wanted to apply it to domains that could make real changes in people’s lives. Emory’s strong healthcare environment made it the perfect place to focus on that.”

Prior to Emory, Yang completed his PhD at the University of Illinois Urbana-Champaign, an institution best known for engineering and computer science. But at Emory, surrounded by hospitals, public health scholars, and NIH-funded faculty, Yang found a new path: developing machine learning tools to advance healthcare—particularly for complex, chronic conditions like diabetes.

Tackling Challenges in AI and Medicine

Despite the promise AI holds in healthcare, Yang points to two major hurdles: data and interdisciplinary understanding.

First, healthcare AI needs real patient data, which is difficult to access, standardize, and process. Unlike traditional lab-generated data, clinical data from electronic health records (EHRs) is messy and often biased, creating risks for misuse or inequity. “If we don’t get the right data,” Yang explained, “the technology won’t help—and could even do harm.”

Second, there’s a gap in knowledge and culture between data scientists and healthcare professionals. “Computer scientists often lack the domain knowledge, and many clinicians don’t yet trust or understand AI tools,” Yang said. “So, collaboration is essential. That’s why I joined EGDRC—to connect with the people who know medicine.”

His advice to students hoping to work at the intersection of AI and healthcare? “Be open-minded and collaborative. The knowledge gap between tech and medicine is real, and no one can work in a vacuum anymore.”

Introducing TACCO: An Algorithm for Personalized Care

Yang’s current work focuses on a novel AI framework called TACCO, short for “Task-specific co-clustering of clinical concepts and patient visits” (published in the KDD 2024 conference). TACCO is designed to do more than just predict health risks—it also identifies meaningful subgroups of patients who share similar risk profiles.

“TACCO uses graph-based models to represent how different risk factors interact,” he explained. These factors—ranging from clinical indicators to genetic and behavioral data—often influence one another in complex ways. Traditional models like logistic regression or random forests can’t capture these interactions, but TACCO can.

The system not only forecasts patient outcomes, but also clusters patients into subtypes with shared risk mechanisms. This allows for more precise treatment recommendations, especially for diseases like diabetes, which affect people differently.

A recent paper presented at the MedInfo 2025 conference extended TACCO to incorporate genomic data, linking genetic risk markers (SNPs) with clinical factors to further refine patient clustering.

What’s Next for TACCO?

Yang’s team is already planning three major expansions. First, they aim to incorporate external knowledge—such as medical literature or knowledge graphs—using large language models. This could reduce the amount of patient data needed while improving prediction accuracy.

Second, they want to broaden the types of data TACCO can analyze. In addition to clinical and genetic information, the system will integrate survey responses, wearable data, and other real-world inputs. Yang is particularly interested in tapping into rich, multi-modal datasets from NIH’s Bridge2AI program and UW’s AI-READI initiative.

Finally, the team hopes to model disease progression over time. “Right now, TACCO only looks at a snapshot—a single hospital visit,” Yang said. “But we need to understand how conditions evolve, especially for chronic diseases like diabetes.”

Collaborating with EGDRC

At EGDRC, Yang says he’s found the kind of collaborative, interdisciplinary environment that makes breakthroughs possible. “Everyone brings different strengths—public health, medicine, computer science,” he said. “That’s what makes this work exciting and impactful. We’re not just building algorithms. We’re working together to improve real lives.”