GenRewrite
Uses large language models and counterexample-guided validation to rewrite SQL queries while preserving semantics.
Self-Improving Data Systems
AI, database learning, and workload feedback for improving queries and data pipelines while preserving semantics and operational constraints.
DAGSmith
Uses pipeline dependencies as optimization signals and separates rewrite generation from equivalence checking.
Research question
How can AI improve queries and data infrastructure while preserving correctness and operational safeguards?
Approach
GenRewrite uses large language models and natural-language rewrite rules to generate efficient, semantically equivalent SQL rewrites. DAGSmith uses pipeline dependencies as optimization signals and separates rewrite generation from equivalence checking.
Key results
- On the open-source Tuva dbt project, DAGSmith reduced elapsed time by 42.6% and warehouse compute cost by 67.7%.
Publications
GenRewrite: Query Rewriting via Large Language Models
SIGMOD 2026 · ACM SIGMOD International Conference on Management of Data
Paper ↗Technical Report ↗Project ↗
Cite
@article{a4a53ccf-7851-464a-8346-dc6c0f5823dc,
title = {GenRewrite: Query Rewriting via Large Language Models},
author = {Jie Liu and Barzan Mozafari},
journal = {ACM SIGMOD International Conference on Management of Data},
year = {2026},
doi = {10.1145/3786684}
}SlabCity: Whole-Query Optimization using Program Synthesis
PVLDB 2023 · Proceedings of the VLDB Endowment
Paper ↗Project ↗
Cite
@article{def837ec-1591-44a4-a36d-929442c789f3,
title = {SlabCity: Whole-Query Optimization using Program Synthesis},
author = {Rui Dong and Jie Liu and Yuxuan Zhu and Cong Yan and Barzan Mozafari and Xinyu Wang},
journal = {Proceedings of the VLDB Endowment},
year = {2023},
doi = {10.14778/3611479.3611515}
}