Authors:
Stephen Day III, Wathila Ekanayake
MIT Supply Chain Management Program (SCM)
Summary:
This project designs a secure, local machine-learning pipeline to extract demurrage and other critical terms from unstructured shipping contracts stored as PDFs and emails, transforming them into a normalized relational database. By combining schema-first design, open-source large language models, and human-in-the-loop validation, the approach aims to achieve extraction accuracy while preserving data confidentiality. The resulting database enables the sponsor company to move from manual, reactive demurrage reviews to proactive, data-driven contract negotiation and transportation cost optimization.
Leave a Reply