Database simulation

My primary research interest is in the efficient simulation of databases connected to various external systems, including local application software and remote LLM-based smart agents. System simulations, also known as what-if analyses, are invaluable for a myriad of applications, from data analytics and future prediction to software tuning, risk management, and correcting past data errors. As relational databases are often used for persistent storage in these systems, efficient database simulation becomes paramount. This is especially true as a system's transaction history expands; replaying transactions without optimization becomes both time-consuming and resource-intensive. My research aims to minimize the overhead of system simulation by developing innovative techniques for efficient database replay while ensuring simulation accuracy.

Multi-modal web agent

Today's data and privacy protection regulations place stringent requirements on the design of data ecosystems, including accommodating the right to be forgotten and adhering to the data minimization principle under GDPR. These requirements, designed to protect individual data rights, have the potential to complicate the circulation and utilization of data. Although a real-life data marketplace must conform to these regulatory requirements, over-stringent data regulations could hamper data trading, which could lead to reduced practicality and social efficiency. Thus, while striving to make our technical mechanisms respectful of laws, it is also crucial to consider designing legal mechanisms in alignment with current technological developments to balance data rights and social welfare. Ideally, combining the design of both technical and legal mechanisms would be invaluable to comprehensively build a practical data ecosystem.