The ASA is built on the strength of its members—their work, expertise, and dedication to the field. This month, we’re highlighting Yuelin Zou, who shares with us details of his career, research, and future goals.
How did you become interested in statistics and data science?
I majored in applied math before pursuing a statistics major. My interest shifted after I met David Draper, former professor of statistics at the University of California at Santa Cruz and an elected ASA Fellow (2007). He had such great enthusiasm for Bayesian statistics that he made me appreciate how elegant and powerful statistical thinking can be: No longer were stats about “numbers,” but rather it was about using statistical reasoning under uncertainty.
Inspired by that, I went on to pursue my bachelor’s degree in statistics at the University of Illinois Urbana-Champaign and later earned my master’s in business analytics from Columbia University. From there, I began my professional career in data science within the tech industry, where I apply statistical modeling and natural language processing to solve real-world business problems.
What is your current role or area of expertise in statistics and/or data science?
In my current position, I focus on developing predictive models through machine learning and generative AI. Most of my work is centered on statistical modeling and time series forecasting, with a focus on using structured data (such as financial and operational metrics) and unstructured data (text and transactions) to assist in the strategic financial decision-making process.
Much of my statistical modeling uses gradient boosting frameworks and time series models to predict various financial, lending, and risk metrics. To ensure the complex models can be interpreted by business stakeholders, I use SHAP [Shapley additive explanations] values, which allow me to ensure the complex models are understandable and explainable to the business stakeholders. In addition to providing transparency into the models, I provide for seasonality, cohort effects, and user growth, which allow for daily-level forecasting.
Overall, my expertise combines statistical learning theory, machine learning interpretability, and financial analytics, enabling organizations to develop data-driven, scalable, and explainable forecasting strategies.
How are you applying GenAI and statistical methods in the finance sector?
For me, the use of GenAI in finance is a matter of achieving an appropriate equilibrium regarding operational efficiencies versus the oversight of risk.
I have been developing a series of large language model–based forecasting and analytical agent systems that leverage internal financial data sets in conjunction with GPT-4 to automate many previously manually and repetitive tasks, such as rolling forecasts, variances, summaries of data, and interactive financial reporting and analysis chatbot systems that enable executives to ask questions of their data in plain English.
To make these systems trustworthy and compliant, I’ve employed various efficient reasoning methodologies, in addition to using supervised fine-tuning on a proprietary set of financial documents. The output from the models is therefore not only contextually correct but also aligns with the data governance practices and compliance regulations of the enterprises.
The GenAI solutions I have developed have demonstrated clear positive effects, enabling executives to act more quickly when making decisions and reducing the amount of repetitive analytical effort required by finance teams, which permits finance professionals to spend more time focused on developing and implementing business strategies, rather than wrestling with data.
What motivates your current research focus on model explainability and unlearning?
I often say GenAI wouldn’t exist without statistics. Statistics provides a means for translating abstract ideas into tangible information and therefore provides a basis for the transition to concrete results. Beyond this, my statistical education has also allowed me to think about uncertainty and causal relationships, all critical skills to develop when you are working with models like large language models.
Currently, researchers continue to study how to get past the “black box” nature of GenAI systems; however, there remains an enormous amount we do not yet know about how GenAI systems retain, recall, and lose knowledge. That’s what drives my research on model explainability and unlearning.
In the context of banking and finance, these questions are especially important. Trust in strategic financial decisions takes years to build, and one opaque model output can ruin everything. We need to know the reasoning behind a model’s prediction, we should try to prevent the likelihood of a hallucination, and we need to provide a way to safely unlearn or modify data based on regulations or fairness. Ultimately, I would like to assist in creating transparent, manageable AI systems that companies will be able to trust to make strategic financial decisions responsibly.
What career advice do you live by, and who gave it to you?
I think the most valuable career guidance I have received from anyone was from my mother. She told me, “Do what you love or, alternatively, if you cannot do what you love, then do what you feel passionate about.” I have been able to carry this concept through all steps of my own journey. New career paths are emerging every year, especially in data and AI. Ultimately, the idea will be to identify an area in which your curiosity exists and pursue it. Curiosity, as I said earlier, when fueled by passion, produces persistence—and it is persistence that will turn your curiosity into expertise.
Are you a member of an ASA chapter or section? Have you volunteered with the ASA in the past? If so, what did you learn from that experience?
I’m currently a member of the ASA New York City Chapter. I recently had the opportunity to serve as a panelist and judge for the ASA x AI4Purpose 2025 Annual AI Workshop, where I gave a talk titled “Benefits, Risks, and Hallucinations of Generative AI.”
Through that experience, I was incredibly inspired by the innovative problem-solving skills of today’s college seniors and recent graduates, many of whom are already building creative AI-driven solutions to real-world problems. Seeing their capabilities made me reflect on my own experiences during my internship in data science, long before GenAI existed, and the time it would take to accomplish what is achievable today in a matter of hours using the right tools.
This experience has further solidified my view that the future of AI has the potential to transform numerous industries, particularly in health care and financial technology, and responsibly innovating within these industries will lead to quantifiable improvements in people’s lives.
Would you like to participate in the Amstat News Member Showcase? Email ASA Communications Manager Megan Murphy and she’ll email you some questions.
Leave a Reply