PRINCE, developed by Bayer and Thoughtworks, is a cloud platform that integrates preclinical safety study reports using agentic RAG and Text-to-SQL. It evolved from keyword search to an intelligent research assistant capable of answering complex questions and drafting regulatory documents.
Bayer AG and Thoughtworks built the Preclinical Information Center (PRINCE), a cloud-hosted platform that integrates decades of safety study reports using agentic RAG and Text-to-SQL. The system evolved from keyword-based search to an intelligent research assistant capable of answering complex questions and drafting regulatory documents.
Preclinical drug discovery is data-intensive, and traditional keyword search (Boolean logic) often fails to address nuanced research questions. Advances in LLMs and RAG enabled combining natural language queries with precise information retrieval. Bayer invested early to leverage these technologies for improving preclinical data access.
PRINCE demonstrates how context engineering (shaping and routing information between specialized agents) and harness engineering (orchestration, recovery, observability) ensure reliability and control. By prioritizing transparency, explainability, and human-in-the-loop integration, the system improves data accessibility and research efficiency while maintaining governance and compliance. This case shows AI's transformative potential in pharmaceutical data challenges.
The comment highlights the underrated aspect of context discipline, noting that larger context windows do not eliminate the need to decide what the model should not see. This suggests that careful selection of input data is crucial for building reliable AI systems.