Rijkswaterstaat Chemical Detector
AI & Data
Semester programme:Enhanced AI Techniques
Client company:Rijkswaterstaat
Project group members:Mbaye,Maimouna M.
Soro,Adama Patrick A.G.Z.P.
Senden,Thijn T.R.A.
Baloum, Nadiem N.
Project description
How can an AI-powered tool automatically extract chemical the necessary information including discharge limits from environmental permit PDFs and match them against a regulated substances reference list accurately and efficiently enough to support Rijkswaterstaat's water quality monitoring?
Context
the domain is Waterkwaliteitsbeheer monitoring and the project uses AI to make the processing of legal-environmental documents faster, more consistent, and scalable.
Results
The project produced three concrete deliverables:
- a fully functional Jupyter notebook pipeline
- a standalone Windows application (.bat) for non-technical users
- a written proposal recommending a more powerful AI model for future use.
The pipeline was tested on real permit PDFs provided by Rijkswaterstaat, demonstrating that the system can successfully extract chemical names, discharge limits and write them to a structured Excel file.