Kun Bu is a postdoctoral fellow at the University of South Florida, researching advanced statistical and AI approaches for extracting trusted knowledge from large and complex datasets. In this Postdoc Portrait interview, she shares how she found new ways to make meaningful connections within datasets to inform better decision-making in science and society.
Improving data integrity with machine learning
Q | What drew you to statistical analysis of large datasets?
I was drawn to statistics and data science because I have always been fascinated by finding patterns in complex systems. When I first started studying finance, I saw firsthand how decisions that affect millions of people are often made under uncertainty. That experience sparked my interest in understanding how data can be used to make more informed and reliable decisions.
As I continued my education, I realized that the same statistical principles could be applied outside of finance. Whether studying drug safety, disease outcomes, human behavior, or financial markets, researchers face common challenges. It’s about extracting meaningful insights from incomplete and often overwhelming amounts of data. I was particularly attracted to the interdisciplinary nature of the field, where mathematics, computing, and disciplinary expertise work together to solve real-world problems.
What keeps me motivated now is the opportunity to develop methods that not only improve predictions, but also help explain why patterns emerge, and ultimately contribute to scientific discovery and better decision-making across the field.
Q | What scientific problem are you trying to solve?
I work to solve a fundamental challenge of the modern data era: how to extract trusted knowledge from large, complex, and interconnected datasets. Many important issues in health, finance, and public policy generate vast amounts of information, but traditional statistical methods often have difficulty understanding the hidden relationships within this data.
My research develops new statistical and AI approaches that can identify meaningful patterns, quantify uncertainty, and support better decision-making. For example, I’ve been working on how to use medical databases to detect drug safety risks. A common theme across my projects is building tools that help researchers move beyond simple predictions toward a deeper understanding of the underlying systems.
Ultimately, my goal is to make data-based discoveries more accurate, easier to interpret, and more useful so that scientists, clinicians, policy makers, and industry leaders can make better-informed decisions in an increasingly data-rich world.
Overcoming misconceptions in AI
Q | What unexpected things have you learned from working with large amounts of data?
One of the surprising lessons I learned is that more data doesn’t automatically mean better understanding. Early in my career as a researcher, I thought that larger datasets would naturally yield clearer answers. Rather, we found that large datasets often introduce new challenges, such as hidden biases, missing information, and complex relationships that can easily lead to misleading conclusions if not carefully analyzed.
I was also surprised by how much uncertainty remains, even with advanced AI and statistical models. The most valuable insights often come not from finding a single “correct” answer, but from understanding the range of possible explanations and quantifying the reliability of the results.
This realization changed the way I approach my research. I’m currently focusing on interpretability, transparency, and quantifying uncertainty, rather than just focusing on building more powerful predictive models. For many real-world applications, helping decision makers understand what we know and what we don’t know can be as important as achieving high predictive accuracy.
Q | If your research is successful, what will change in science and society?
This effort could help make data-based decision-making more reliable, transparent, and accessible across many sectors. As the datasets that organizations collect become increasingly large and complex, there is a growing need for methods that not only produce accurate predictions, but also provide meaningful explanations and quantify uncertainty. In the medical field, this has the potential to detect drug safety risks early and improve patient outcomes. More broadly, the statistical and AI methods I have developed can help researchers discover hidden patterns in complex systems and transform data into actionable knowledge. Ultimately, I hope that my work will contribute to a future where data is used not just to automate decision-making, but to enhance human understanding and support more thoughtful, evidence-based choices in science, industry, and society.
Q | What question do you want to answer next?
I’m most excited about answering how AI and statistical inference can be combined to better understand complex systems rather than simply predicting outcomes. Although AI has achieved remarkable success in pattern recognition, many models still operate as “black boxes”, making it difficult to understand why they make certain predictions or how much confidence they should have. I am particularly interested in developing methods that integrate the predictive power of AI with statistical interpretability and uncertainty quantification. By doing so, you can create tools that not only identify patterns within large datasets, but also reveal the underlying relationships that drive those patterns. Whether in healthcare, finance, or other scientific fields, answering this question can help researchers and decision makers move from prediction to understanding, and potentially enable more reliable, transparent, and actionable use of data when solving real-world problems.
Answers have been edited for length and clarity.
Are you a researcher who would like to be featured in the “Postdoc Portraits” series? Submit your application here.
