Enterprise Text-to-SQL evaluation
Led IBM’s research on enterprise Text-to-SQL evaluation and released the open-source IBM Text-to-SQL Evaluation Toolkit. Related work was presented at ICDE and VLDB.
Evaluation for Data & AI / 2007–present
Benchmarking Done Right
Evaluating AI systems for the work they need to do.
AI systems make evaluation more important—and more difficult. A benchmark score alone does not establish whether a system will work on enterprise data. I develop evaluation methods, tools, and shared benchmarks, drawing on research in clustering, Text-to-SQL, semantic table interpretation, and query processing.
Seven years of SemTab leadership
Co-founded and co-organized the SemTab challenge at ISWC from 2019 to 2025, developing shared evaluation resources for semantic table interpretation.
Evaluation across data systems
Research on clustering evaluation, data discovery benchmarks, tabular representations, and variation between repeated semantic query executions.