GUEST COLUMNISTS

• You Can't CAPA What You Can't Classify

  An LLM or agentic error can occur with no component broken in any way the QMS recognizes. Classifying it takes two fields: the error type and the error origination point.

• 6 Modeling Approaches That Make Complex Bioprocess Data Useful

  AI algorithms can reduce data dimensionality more faithfully by interpreting multivariate features relevant to prediction or control.

• Building Trust In AI For Computerized System Validation

  Using AI for CSV might feel like the students are grading their own papers. One consultant explains why it's not the same, namely, there are no subjective judgment calls.

• A Gentler Cell Line Selection Method Using microRNA

  Traditional selection stresses surviving cells. New research shows how structurally interacting RNA may use endogenous microRNA for a faster, nontoxic selection process.

• Controls For Retrieval-Augmented Generation In GxP

  Validation teams deal with inherent variability all the time. The same principles for GMP — think visual inspection — can apply to probabilistic AI systems.

• Implementing AI Solutions In GMP While Managing Risk

  When using artificial intelligence in GMP settings, risk management extends far beyond  validation. This discussion explores moving AI solutions out of the pilot phase.

• Implementing And Optimizing Your eQMS

  This article addresses everything that follows the signed contract for an eQMS: how a system is configured, validated, adopted, and maintained.