Meet Aaron Mullen

“My experience at CAAI has profoundly shaped my professional goals. Its focus on innovation in AI and data science directly motivated my decision to pursue a Master’s degree and has provided me with invaluable, continuous learning opportunities.” 

Aaron Mullen took his first programming class in high school, but was interested in data science before knowing what it was. He was interested in interpreting and visualizing patterns and trends. After learning about data science as a field, he knew it was something he wanted to pursue. In 2022, Aaron began working in the department that would become CAAI as an undergraduate intern. He worked through his junior and senior years until becoming a full-time staff member upon graduation. He is now pursuing a Master of Science in Data Science.  

Throughout his time at CAAI, he has worked in a variety of fields, such as website development, machine learning (ML), and data visualization. He specializes in and has worked on projects in data science and ML methods like classification and time series forecasting. He has contributed to multiple CAAI projects, such as CLASSify, an online tool for easily performing ML classification and data analysis created by Aaron. He also works with the Rapid Actionable Data for Opioid Response in Kentucky (RADOR-KY) team to build forecasting models to predict future opioid overdose trends for different areas of the state of Kentucky. 

1. What prompted you to develop CLASSify, and what would you like to see future versions be able to do? 

Classification problems are one of the most common types of research questions, but using machine learning to solve these kinds of problems requires a lot of background knowledge. We knew we wanted to make these machine learning methods accessible to everyone, so we developed CLASSify as a way to allow researchers to easily upload their own data and quickly gain insights into their specific problem. We’ve expanded CLASSify a lot since we started it to include features like synthetic data generation, unsupervised clustering, and feature importance scores, but there is still more we plan on adding. The vision is to make CLASSify a hub for all kinds of tabular data analysis, so we want to improve it to allow users to analyze and explore more kinds of data through natural language questions, and modern AI techniques can answer those questions with explanations and visualizations. 

2. Have there been any unique struggles or obstacles you’ve faced in your work with data and AI? What makes these things difficult? 

Machine learning can be a very daunting field to approach because the models themselves can be very difficult to interpret and understand. People have a natural tendency to want to know what’s going on “under the hood” if they want to trust the output of these machine learning models. But while many of these models, such as neural networks, are designed to be efficiently and effectively used by computers, they are much less understandable from a human perspective. Still, more and more techniques are being invented to provide explainability to machine learning techniques, which allows the results of these models to provide more useful insights into data trends and relationships. 

3. What do you find most interesting in terms of how AI can be used to process or interpret data? 

Because of how differently machine learning works from traditional human techniques to understanding data, it can capture relationships in data that may not have been identified otherwise. This allows for more advanced and complete interpretations of data, and I think the combination of human and machine analysis can create the best results, as both can account for the weaknesses of the other. 

4. If you could take on any challenge or goal with your work, what problems would you love to try and solve?  

I am most interested in data visualization and how it can be used to provide insights into data analysis and machine learning. I would love to try to explore how visualizations can provide more interpretability for how machine learning models work and how they reach results. Also, the concept of automating the creation of visualizations on an interface like CLASSify so that graphs can be generated real-time to answer user questions for new data seems very interesting to me. 

CAAI team members include high school, undergraduate and graduate students, as well as entry-level and seasoned professionals. The Center thrives by fostering collaboration between early-career and late-career professionals, creating a dynamic environment that promotes a culture of experimentation, exploration, and career development.