IRisk Lab Fall 2026 Projects

The IRisk Lab will establish a strategic collaboration with the Illinois Department of Insurance to advance research on emerging challenges in insurance regulation, risk management, health insurance, and artificial intelligence. Through jointly developed research projects, policy analyses, and data-driven solutions, the partnership will address real-world issues and support evidence-based regulatory decision-making.

A central objective of this collaboration is to provide students with meaningful experiential learning opportunities by working alongside regulators on real-world projects. Students will gain firsthand exposure to the regulatory environment, develop practical skills in insurance policy and analytics, and better understand how quantitative methods inform public decision-making.

This unique partnership will bridge academic research and public service, preparing the next generation of actuaries, data scientists, and risk professionals with both technical expertise and valuable experience in the regulatory sector.

Supervisor: Frank Quan

Large language models (LLMs) are rapidly transforming everyday workflows by improving productivity, automating routine tasks, and expanding access to information. However, their performance often declines when they are applied to specialized insurance tasks that require domain expertise, numerical reasoning, regulatory knowledge, and the accurate interpretation of technical terminology.

This project aims to develop a domain-adapted, open-source LLM specifically for insurance applications by retraining and fine-tuning foundation models using curated insurance data and actuarial knowledge. Particular emphasis will be placed on improving numerical calculation, quantitative reasoning, policy interpretation, and other tasks that require a high level of accuracy and reliability.

Beyond model development, we will design and deploy an extensible AI toolkit that integrates the specialized LLM with domain-specific tools and workflows. This toolkit will support practical applications such as underwriting assistance, claims analysis, regulatory compliance, actuarial calculations, and decision support.

The resulting platform will provide an accessible, transparent, and customizable AI solution that advances insurance research, education, and professional practice while reducing reliance on proprietary, closed-source systems.

Supervisor: Frank Quan

Crypto markets can move very quickly. When a large trade enters the market, it may push the price up or down, especially when there is not enough liquidity available at the best prices. In traditional financial markets, large orders are often split into smaller pieces to reduce price impact. In crypto markets, this issue can be even more important because trading volume are high, while the actual executable depth of the market can be relatively thin.

A key example is a liquidation cascade. Many crypto traders use leverage, meaning they borrow or use margin to increase their trading positions. When prices move against them, their positions may be automatically liquidated. These forced trades can push prices further in the same direction, which may then trigger additional liquidations. As a result, an initial price movement can turn into a much larger market crash. For example, on October 10, 2025, CoinDesk Data described the largest crypto liquidation event in history: about $19 billion in notional positions were unwound within 24 hours, total open interest fell by 27.5%, and almost $60 billion of open interest disappeared, with the steepest drawdown concentrated within about 25 minutes.

In this project, we will study how large trades and liquidation events affect crypto market liquidity. We will compare the cost of executing a large order all at once with strategies that split the order into smaller pieces over time or across venues. We will also analyze when liquidation cascades are most severe, such as during periods of low liquidity, high volatility, heavy leverage, or exchange-specific market stress.

Throughout the semester, we will collect and analyze crypto market data, measure price impact, identify periods of liquidity stress, and evaluate whether better execution design can reduce trading costs and market instability. The project will connect crypto market behavior to broader questions in finance and risk management, such as how trading platforms should manage liquidity risk and how market design can reduce the chance of sudden price crashes.

Supervisor: Xiaochen Jing

Graduate Supervisor: Zhonghe Wan

This project is co-hosted with the Department of Climate, Meteorology, & Atmospheric Sciences. It bridges climate science, actuarial science, data engineering, and machine learning to address urgent catastrophic risk challenges. Students will explore how hail hazards translate into physical damage and financial loss, with emphasis on real-world catastrophe risk modeling. The program also immerses students in the full research lifecycle: collecting data, engineering multi-source inputs, building predictive models, validating results, and interpreting risk insights.

Project description

ASRM Supervisor: Frank Quan

CliMAS Supervisor: Robert Trapp

Graduate Supervisor: Jiayi Guo