Physiologically based biopharmaceutics models (PBBM) link drug properties to in vivo performance, but building them is manual and time-intensive. Translating a formulation question into a decision typically requires weeks of simulation, parameter sourcing, and documentation. This session introduces a modeling-in-the-Loop framework that couples large language models (LLMs) with a deterministic PBPK engine (Gastroplus) for natural-language model execution and interpretation.

Key Highlights

• Introduction to Agents: Opens the black box by explaining LLM fundamentals, agent architecture, and how prompting, context, and tool access shape outputs.
• Practical Modelling Workflow: Translates a formulation question into live simulations using agents, covering how to build scenarios and critically analyze agent-generated responses. This uses the GastroPlus® engine as the sole numerical truth. LLMs act strictly as orchestrators, routing natural-language intent.
• Defined Boundaries: Clearly distinguishes where agentic AI excels (schema extraction, parameter mapping, scenario reasoning) versus where human judgment remains non-negotiable (mechanistic structure, physiological interpretation).

Presenters:

• Priyata Kalra – Senior Principal Scientific Product Manager
• Neil Miller – Vice President, Product
• Becky Graves – Senior Director, Scientific Product Enablement