During my PhD at EPFL, I am currently working on algorithms for query optimization in database systems. In particular, I exploit structural information of the queries to help reason about their theoretical worst-case asymptotic complexity and guide query optimizers towards good execution plans in practice.

In the case of conjunctive (select-project-join) queries, I proposed a query optimization approach based on "meta-decompositions", a representation that succinctly encodes the structures of all possible join trees for acyclic queries and allows for an efficient dynamic programming algorithm for cost-based optimization. Not only can this representation be viewed as a helpful strategy to guide cost-based optimizers to efficiently find query plans that are likely good, but it can also be helpful for many current theoretically desirable structure-guided approaches that require enumerating or selecting optimal join trees.

I also extended similar ideas to recursive queries, in, e.g., Datalog and Recursive SQL, and proposed the notion of size bound–adorned datalog which allows us to derive, for the first time, a series of theoretical upper bounds on asymptotic complexity and result sizes of recursively defined relations.

Earlier, during my internship at EPFL, I worked on algorithms to efficiently approximate results of aggregation or rollup queries, based on partially materialized data cubes that contain aggregation results projected to lower dimensions.

During my undergraduate study of software engineering at McGill, I worked on model-driven software engineering, especially in the context of test case generation for autonomous vehicles. We designed a system Concretize which automatically generates concrete test scenarios (that can be run directly in simulators) based on high-level specifications by the user in a domain-specific language.

We further proposed our vision on a collaborative "scenario catalog" that facilitates sharing and analysis of test cases and results, based on a unified metamodel of test scenarios and related statistics.