A Tandon computer engineering student lends her LLM expertise to a quant competition

When the Society of Quantitative Analysts (SQA) held its 2024 Alphathon competition this fall, one of the prize categories involved how to use Large Languages Models (LLMs) in investing — among one of the hottest topics in FinTech.

The month-long event drew everyone from graduate students to professionals with decades of experience, and at the finals, held in New York City in October, one clear winner in the LLM category emerged: a team calling itself Big Red, Big Purple, whose solution was a 2-D Bayesian regime switch that successfully outperformed the S&P 500.

The “Purple” in the team name was an homage to NYU Tandon computer engineering master’s student Tyler Li (‘26) who joined a group of friends studying at Cornell to enter the Alphathon. “Their passion for the topic really sparked my curiosity, especially when it came to exploring how LLMs could support investment analysis,” she explains.

She had met two members of the group, Elina and Hudson, while studying for their undergraduate degrees at Queen’s University, in Kingston, Canada. While they ended up embarking on separate paths for graduate school — Elina and Hudson choosing to maximize their quantitative skills at Cornell, with Tyler choosing NYU to dive deeper into the tech — they remained close. 

Their main takeaways from the competition project, which was sponsored by AllianceBernstein, included:

• LLMs are good at market sentiment analysis but require care in their use
• Using Gaussian Mixture Model to classify market trending is effective
• Using Bayesian Inference and Mean Variance Optimization can help find the best strategy

“It was exciting to apply what I had learned from research papers and blogs to a real-world challenge,” Tyler, who has interned at such major companies as Microsoft and DiDi (a ride-hailing app popular worldwide), says. “The experience showed how LLMs can support strategies, rather than fully replacing them. It also gave me a clearer understanding of the current limitations of LLMs and the direction we can take to use them more effectively in enhancing existing work and strategies.”

Li’s teammates

• Hudson Chen
• Joshua Ma
• Yuao Peng
• Shun Wang
• Elina Zhuang