Modern data-intensive applications rely on graph databases, yet graphs frequently contain inconsistencies that violate structural constraints. My work addresses this from a human-centric perspective: rather than fully automated fixes, I develop interactive repair frameworks that keep users in control of the process.
Key research questions#
- How can we design repair systems that are transparent and explainable to non-expert users?
- How can constraint violations be detected and resolved interactively, at scale?
- What role can Large Language Models play in assisting constraint mining and repair?
Selected work#
- Grafixer — a system enabling user-centric repairs for property graphs, presented as a demo at SIGMOD 2025.
- User-Centric Property Graph Repairs — a full framework for interactive repair of property graphs, published in SIGMOD 2025.
- Interactive Graph Repairs for Neighborhood Constraints — introducing neighborhood-based constraint repair at EDBT 2024.
- Graph Consistency Rule Mining with LLMs — using LLMs to assist in mining graph constraints, at EDBT 2025.
- What If: Causal Analysis with Graph Databases — causal analysis framework for graph data, at VLDB 2025.