CFEM Student Recognized by INFORMS 2024 Winter Simulation Conference

Students at Cornell's Financial Engineering program have many opportunities to engage in quantitative finance research and to expand their knowledge through collaborations with industry practitioners and academic experts. This emphasis on a hands-on approach emphasizes practical applications of theory and enables students to explore advanced topics that bridge complex models with real-world solutions. A case in point is the research of current student Daham Kim '24, who, in collaboration with CFEM Visiting Faculty and longtime collaborator Irene Aldridge, has received recognition from the INFORMS Winter Simulation Conference.

Daham's project began as a conversation with Aldridge about various simulation methods in financial markets. This discussion led them to consider the potential of neural network models to address gaps in financial simulations, specifically in capturing the unpredictable dynamics of market scenarios. Their focus turned to attention-based AI models, which have shown notable effectiveness in understanding complex data patterns. While many models have attempted to predict financial time series data, no prior research had simulated these series under diverse market conditions using advanced transformer models.

Aldridge explains, "We realized the potential of such a combination and set out to develop a new quantitative financial framework, which can incorporate diverse scenarios generated by Large Language Models (LLMs)." The team harnessed LLMs to generate extreme market scenarios, capitalizing on AI's ability to "hallucinate" hypothetical conditions. Using transformer models like the Temporal Fusion Transformer, they successfully simulated financial market patterns induced by these extreme, LLM-driven scenarios. Their innovative approach has demonstrated performance advantages over traditional Monte-Carlo simulation methods across several metrics.

Reflecting on the experience, Daham noted, "It was great to see diverse technical expertise and different research interests on campus, creating meaningful research!" This groundbreaking work has been accepted for presentation at both the 2024 INFORMS Winter Simulation Conference in Orlando, Florida, and the 5th ACM International Conference on AI in Finance.