Victor Bechmann and Darius Rohani

Validating LLMs for GMP: A Framework for Document-Centric Use Cases

Large language models (LLMs) are rapidly expanding the possibilities for automation in GMP-regulated environments, particularly in the review and interpretation of unstructured data. This presentation introduces a practical, risk-based framework for validating LLMs used in document-heavy pharmaceutical processes. It will be based on the article: “Seven Control Layers for Large Language Models in GMP Decision-Making”, published in the Jan/Feb 2026 GAMP issue. 

The focus is on use cases where LLMs assist in reviewing and extracting insights from a broad range of data types - including free text, drawings, technical specifications, and structured forms - whether originating from digital or paper-based sources. Common applications include the assessment of batch records, Certificates of Analysis, equipment documentation, and other critical GMP records.

Key topics will include:

• Risk-based validation strategies for LLM-based systems
• Defining appropriate performance metrics for mixed-content data
• Human-in-the-loop controls to ensure accuracy and compliance
• Considerations for traceability, auditability, and change management

Attendees will come away with a clear understanding of how LLMs can be responsibly integrated into existing GMP processes to reduce manual workload, improve consistency, and maintain regulatory readiness—while ensuring that AI implementations align with established quality and compliance expectations.

Bio:

Victor Bechmann: Pharmacist and background as QA/QP. 
Specialist in AI in GxP Compliance at Kuatro Group. 
Author of the article Seven Control Layers for Large Language Models in GMP Decision-Making”, published in the Jan/Feb 2026 GAMP issue. 
He has been validation lead on one of the only full validated AI Vision System for GMP Decision Making in Europe.

Darius Rohani is a Partner at Kuatro and Senior Data Scientist specializing in generative AI, machine learning, and large-scale data solutions. He has led the development of AI-driven systems in compliance-critical environments, translating complex business needs into scalable technical architectures.
Darius was recognized as a Top 100 Talent in Data Science (2022) and brings a strong background from both academia and industry, combining deep technical expertise with strategic advisory experience in enterprise AI transformation.