TRL Lab

Welcome to the page of the Table Representation Learning (TRL) Lab ✨!

Our research is focused on representation learning and generative models for tabular data, such as relational tables. While most AI research attends to images and text data, proportional progress on tabular data is lacking. Tabular data, however, is prevaling in the data landscape and contains highly valuable information that powers important decisions in, for example, governments, enterprises, and healthcare.

In this context our research focuses on topics such as new table embedding models, table retrieval mechanisms, proactive dataset search tools, data semantics, insight retrieval from relational databases, neural predictive tabular ML, agentic data science, and more. Find an overview of published work here. Generally, our research spans machine learning, databases, information retrieval, natural language processing, and human-computer interaction.

The TRL Lab is led by dr. ir. Madelon Hulsebos, and hosts an excellent group of postdoc, PhD, and MSc-level researchers at CWI. The lab is also affiliated with the Table Representation Learning research theme at ELLIS unit Amsterdam, where we organize a seminar and reading group. Beyond CWI, we currently collaborate with researchers at the University of Amsterdam, UC Berkeley, the UN Humanitarian Data Centre, and Google Research. The lab is currently funded by an AiNed grant from the Dutch Research Council (NWO), an industry gift from SAP, and several compute credit grants from OpenAI, Google, and Cohere.

Check our open positions to join us! For research- or societal collaborations, please reach out.