Large-scale Intelligent Data Systems (LIDS) Laboratory

Projects

• Relational Deep Learning

  Machine learning directly over multi-table relational databases, made efficient with SQL and schema information.

• Graph Neural Networks

  Scalable training of graph neural networks on large knowledge graphs.

• Infrastructure for Data Agents

  Giving LLM-based agents reliable access to data sources.

Recent publications

• STEER: Reducing Inference Cost in Relational Foundation Models through Semantically Informed Sampling

  Abdalla Mohamed and Ashraf Aboulnaga

  arXiv:2610.00907. October 2026.Under review.

  • arXiv
  • Code

• From Transport to Grounding: Designing MCP Connectors for Reliable SPARQL Querying by LLMs

  Talha Tahmid, Humera Sabir, and Ashraf Aboulnaga

  VLDB 2026 Workshop: Tabular Data Analysis (TaDA), Boston, Massachusetts, USA. September 2026.

  • Published version
  • Code

• MetaSieve: Faster Relational Deep Learning through SQL-Based Metapath Selection

  Fahim Shahriar Khan and Ashraf Aboulnaga

  arXiv:2608.25903. August 2026.Under review.

  • arXiv
  • Code

All publications