I study database systems, with a focus on query execution and transaction processing. I am interested in how multi-writer databases route transactions that access the same data. My current research explores shared memory and semantic validation for data agents over evolving data.
Research projects
Shared memory for data agents
MAVRA · 2026–Present
I built MAVRA in Rust to let data agents share versioned definitions of metrics and features in PostgreSQL. The prototype tracks whether each definition remains valid after a data update and checks which revision a query uses.
In a collaboration between RUC’s database systems group and Kingbase, I implemented a workload-aware router in C++. The router uses METIS to partition a graph of transactions and the data they access. It directs transactions with overlapping access patterns to the same node to limit cross-node communication.
Industry collaboration · C++
Query execution, indexing, and recovery
RushDB · 2024–2025
My RushDB teammates and I implemented B+ tree indexes, concurrency control, and log-based recovery in RucBase. For the OceanBase Database Competition, we extended MiniOB to handle nested and correlated subqueries and optimized vector search in OceanBase.