I build reliable AI systems, from document understanding and multimodal models to agents, evaluation, and production ML. Through Probably Approximately Human, I help teams turn research into practical products and ambitious AI roadmaps.

I completed my PhD at KU Leuven in 2024, with research on AI-driven document understanding, predictive uncertainty, and calibration. My industry experience includes Instabase and seven years at Contract.fit, bringing together applied research and hands-on product engineering.

My new portfolio brings together my current projects, publications, talks, and consulting services.

Research archive

This site preserves my earlier research, writing, and academic activities. Existing paper, slide, and thesis links remain available.

• Research, publications & thesis
• Blog: Intelligent Automation
• Academic activities & supervision
• News archive

PhD: Intelligent Automation for AI-driven Document Understanding

Completed at KU Leuven in 2024. My thesis explores reliable document understanding through uncertainty estimation, document classification, and the DUDE benchmark.

Read the thesis (PDF) · Public defense slides (PDF) · Project spotlight video

Contact

Email me · LinkedIn · GitHub · Google Scholar