Madelon Hulsebos

Madelon Hulsebos is a Researcher at CWI in Amsterdam, where she leads the Table Representation Learning (TRL) Lab and contributes to the Database Architectures group. She is also a faculty member of the European Laboratory for Learning and Intelligent Systems (ELLIS) Amsterdam unit. Her career bridges academia and industry, including a postdoctoral fellowship at UC Berkeley and prior industry experience in automating data analysis pipelines with ML.

• Education: PhD in Computer Science (University of Amsterdam, 2023), with research at Sigma Computing and MIT; Postdoctoral Fellow (UC Berkeley, 2024).

Her research focuses on establishing tabular data as a key AI modality through Table Representation Learning, generative models for relational data, and robust systems for data analysis. Key interests include:

• Relational Table Embeddings
• LLMs for QA/text2SQL and data wrangling
• Retrieval over Data Lakes and Databases
• Agentic Systems for Data Science
• Democratizing insights from structured data

Recent work highlights trends in benchmarking table retrieval (TARGET), semantic column detection (AdaTyper, Sherlock), and large-scale tabular data curation (GitTables, SchemaPile). These projects address challenges in metadata utilization, data lake search, and end-to-end systems for structured data.

She has secured significant funding, including the NWO AiNed Fellowship Grant ($1M) for her 5-year DataLibra project. Madelon organizes workshops at NeurIPS, SIGMOD, and ACL, and reviews for top venues like VLDB and NeurIPS.

Scientific Awards:

• NWO AiNed Fellowship Grant ($1M)

She actively mentors students and collaborates on European AI initiatives, including monthly TRL seminars and workshops. Her lab's tools (GitTables, TARGET) are widely adopted for training foundation models on tabular data.