Scholar
Madelon Hulsebos
Google Scholar ID: 6IWQn2EAAAAJ
Research faculty, CWI
Citations & Impact
Citations
1,047
H-index
13
i10-index
14
Publications
20
Co-authors
19
list available
Publications
Resume (English only)
Academic Achievements
- Created large-scale datasets: GitTables (1M+ real-world tables) and SchemaPile (real-world DB schemas)
- Developed Sherlock for semantic type detection in tabular data
- Proposed the TARGET benchmark for table retrieval in end-to-end querying
- Built DataScout for proactive, task-based dataset search
- Conducted research on table embeddings (e.g., Observatory project) and role of embedding metadata in retrieval
- Work on contextual sensitive data detection and LLM robustness for tabular reasoning
Research Experience
- Postdoctoral researcher at UC Berkeley
- Over 2 years of industry experience developing ML-driven automated data analysis tools prior to academia
- Initiated TRL efforts since 2020, including founding the TRL workshop series at NeurIPS and ACL
- Established the TRL research theme at ELLIS Amsterdam and coordinates related activities
- Serves as reviewer for conferences/workshops including VLDB, SIGMOD, NeurIPS, and ICLR
Background
- Faculty member at CWI, leading the Table Representation Learning (TRL) Lab
- Member of the Database Architectures group at CWI and faculty at ELLIS Amsterdam
- Research focuses on Table Representation Learning (TRL) and generative models for tabular data
- Aims to democratize insights from structured data and establish tabular data as a core AI modality alongside images and text
- Research supported by NWO AiNed grant, BIDS-Accenture Fellowship, and industry sponsors
