Researchers have developed a novel agentic-AI platform designed to transform fragmented technical information from Chemistry, Manufacturing, and Controls (CMC) process development into a structured, queryable knowledge graph. This platform creates a dual-layer graph: a base lexical layer for document hierarchy and an intelligence layer for ontology-aligned entities and cross-document concepts. LLM agents utilize these layers to answer questions, with a router agent selecting the optimal retrieval path. The system was evaluated on Sanofi data, achieving 95% accuracy on multiple-choice questions and 85% on stricter LLM-judged assessments, revealing a failure taxonomy not captured by simpler metrics. AI
IMPACT This approach could significantly improve knowledge management and traceability in the pharmaceutical industry, accelerating drug development and regulatory processes.
RANK_REASON The cluster describes a research paper detailing a novel AI platform and evaluation protocol.
- alphaXiv
- CatalyzeX Code Finder for Papers
- Connected Papers
- Coordinating Multiple Conditions
- DagsHub
- Gotit.pub
- Hugging Face
- Litmaps
- retrieval-augmented generation
- Sanofi
- ScienceCast
- scite Smart Citations
- CatalyzeX
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
