Harry Reyes Nieva: Human-Centered AI for Health Care

Back in high school, Columbia scientist Harry Reyes Nieva, PhD, was hardly a contender for “Most Likely to Go into Artificial Intelligence” Though Reyes Nieva was a math whiz growing up in Brooklyn and he graduated high school at the age of 16, his family didn’t own a computer until his mid-teens. He planned to become a lawyer.

“I was one of the first in my family to attend college and when I started, I fell into the classic trap of thinking there were only a narrow set of successful career options: lawyer, doctor, and working in business,” recalls Reyes Nieva, who was born in New York and raised in a family originally from Puerto Rico and Cuba. “In my mind, science and medicine weren’t viable paths. Becoming a lawyer seemed more realistic. I didn’t know any scientists and had never met a doctor with ties to Latin America.”

The first major inflection point in Reyes Nieva’s trajectory from aspiring attorney to AI researcher in medicine, was a stint in health care advocacy. His work with health care nonprofits during college helped steer him toward medicine, but his financial aid fell through soon after he arrived in Boston to start a post-bac program to take the prerequisite courses for application to medical school.

“I had to scramble and find a job. I went to work as a research assistant in the general internal medicine division at Brigham and Women’s Hospital, which just happens to be a leader in medical informatics,” Reyes Nieva says, “and it changed my life.”

Making sense of chaos

At the Brigham, Reyes Nieva supported research on influenza and antibiotic stewardship in primary care, then moved on to work for the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) in Nigeria, managing data quality and designing informatics tools that helped providers make clinical decisions.

With his deep expertise in informatics and extensive real-world experience, he joined Columbia’s Department of Biomedical Informatics as a graduate student in 2019, while maintaining an appointment as a visiting postgraduate fellow at Harvard Medical School, motivated to better leverage the tremendous potential of AI and informatics for medicine and public health. Since completing his PhD in Fall 2024, he has been a postdoctoral research scientist in the Department of Medicine’s infectious diseases division and was an AInet Fellow for Safety and Security in AI through the German Academic Exchange Service, Deutscher Akademischer Austauschdienst. 

Today, Reyes Nieva describes his work as developing and applying responsible, human-centered AI approaches that realize the promise of AI to advance science and human health while also examining and addressing the perils of AI for harm, in large part by designing AI with these concerns in mind and ensuring proper oversight of AI tools once they’re deployed.

One of the big challenges in the field is how to make sense of data coming from multiple different sources when that data is also continuously changing.

“Harry's work is so important because it turns that chaos into useful insights about our patient population and their care,” says Reyes Nieva’s thesis adviser, Noémie Elhadad, chair of the Department of Biomedical Informatics at the Vagelos College of Physicians and Surgeons (VP&S).

Those insights can ultimately improve health outcomes, sometimes immediately. In one instance, Reyes Nieva’s work has provided a near-real-time picture of infectious disease across New York City by tapping into data from nearly 5 million electronic health records spanning multiple health systems. “Traditional reporting systems are built for an earlier era, leading to delayed case reports, siloed laboratory feeds, and ultimately paint an incomplete epidemiologic picture” Reyes Nieva says. “With our data pipeline and surveillance approach, we were able to identify existing gaps in routine screening that could help public health agencies be more responsive, timely, and precise in their actions.”

By synthesizing vast amounts of scientific literature and national public health and insurance data, Reyes Nieva’s methods have also identified gaps in medical knowledge that need to be filled to improve health care in the long term. “More than 39 million scientific articles are stored in PubMed, the largest biomedical literature repository, and mining that much information is incredibly difficult. It can only be fully leveraged with AI and informatics tools,” Reyes Nieva says. “My approach uses a number of different techniques, including machine learning, natural language processing, and knowledge engineering to give us a sense of the populations, clinical conditions, and topics that are well-represented and those that are not.

“The goal isn’t just to characterize what’s present in the literature—which is hard enough to do—but to illuminate uncharted areas of scientific inquiry so we can potentially drive research in those new directions and ensure that there's equitable coverage across different populations and different conditions.”

AI and ID

After finishing graduate school, Reyes Nieva joined the infectious diseases division as a postdoctoral research scientist working formally under division chief, Magdalena Sobieszczyk. “I love ID medicine and have worked closely with many ID doctors throughout my PhD and at PEPFAR. Jason Zucker has played an especially integral role during my time at Columbia. More recently, I’ve also begun collaborating closely with epidemiologist and implementation scientist, Delivette Castor, who has been a great support. Now that I’ve joined ID, my goal is to bring my particular areas of expertise to bear to advance scientific discovery and health care in this space. I feel like I’ve been slowly building my skill set, working my whole career toward this new home, and all the pieces are finally falling into place.”

Among his many projects, Reyes Nieva is teaching informatics to the division’s clinical fellows, developing AI tools to better understand a variety of diseases including long COVID, and creating specialized AI agents that collaborate to solve complex problems by dividing tasks and coordinating actions to help guide patient care in real time, perform disease surveillance, and collaborate with physicians and patients to improve understanding and shared decision-making.

“Dr. Reyes Nieva adds a unique and exciting dimension to our work in the division,” says Sobieszczyk. “The full potential of AI has not been completely recognized in infectious disease, and Harry’s expertise in AI, informatics, and public health will add precision and innovation to infectious disease surveillance, as well as improve prevention, care, and treatment of key issues like HPV, STIs, and long COVID.”

It’s important to Reyes Nieva that these tools are sustainable—that is, they continuously adapt as conditions evolve. “As conditions on the ground change, we need to update our models, understanding, and processes accordingly. It's not just a one-and-done thing; it’s essential that we constantly monitor and update our systems and approaches, otherwise we run the risk of poor utility and unintended consequences.”

Currently, Reyes Nieva sits on the governance and oversight working group of the AI at VP&S Initiative to provide guidance and help create structures to continually scrutinize AI tools deployed at CUIMC. He also participates in the AI at VP&S community working group to help ensure the use of AI at CUIMC is informed by the people and patients who are impacted by it.

“At the end of the day, my reason for pursuing AI in medicine and public health is the same reason I wanted to become a lawyer: to help people,” Reyes Nieva says. “I believe with the advent of this technology, we now have tremendous opportunity to improve human health and wellness and do it at an unprecedented scale.”