Regulatory Document Monitoring

Mission

Support for regulatory surveillance teams through the integration of intelligent agents and GenAI to automate and secure the comparison of Summaries of Product Characteristics (SmPC) based on ATC codes. Industrialization of pharmaceutical monitoring: automated extraction, structuring and analysis of large regulatory document sets. Development of agentic AI agents (beyond LLMs) orchestrating complex tasks: contextualized querying via Retrieval-Augmented Generation (RAG), automated synthesis of differences between molecules, proactive anomaly detection, dynamic notifications and autonomous initiative-taking for supervision. Intelligent vectorization and advanced knowledge structuring through embeddings on a regulatory corpus, with optimized storage in vector databases for contextualized search and reasoning. Implementation of analytical decision-support algorithms, anomaly detection and discrepancy scoring to strengthen the reliability of critical document analyses. Deployment of interactive dashboards for dynamic reporting of anomalies and AI recommendations, with automated report generation for decision support and operational guidance. Advanced automation through object-oriented architecture: modular prompt generation, complex reasoning chains (multi-step planning), tuning of agentic orchestration to maximize relevance and operational safety. Robust testing (Pytest), continuous agent monitoring (retrieval/generation metrics, Langfuse) to ensure performance, explainability and transparency.

Contexte & Environnement

Technological modernization of ANSM in pharmacovigilance and document regulation, led by a consortium (Theodo x Cleyrop).

Objectifs Cles

Technologies & Infrastructure