AstraZeneca launches Research Assistant for AI-driven drug discovery

AstraZeneca has developed *Research Assistant*, an internal agentic system powered by large language models (LLMs) to streamline drug discovery and clinical research. The platform aggregates data from scientific literature, knowledge graphs, chemistry databases, clinical trials, safety resources, and internal experimental systems into a unified chat interface. Users can toggle between a fast mode for immediate answers or a multi-step mode for complex research tasks, with all responses grounded in retrieved evidence and linked to original sources for traceability.

The system, detailed in a 16-page paper (arXiv:2608.12395), was deployed at scale across AstraZeneca’s R&D teams to support day-to-day workflows. The architecture emphasizes modularity, allowing integration of proprietary and public datasets while maintaining auditability through source attribution. The team behind Research Assistant—led by Piotr Grabowski and 18 co-authors—highlighted lessons learned from scaling the system, including challenges in data heterogeneity and the need for robust evidence retrieval.