Longer-form work

From Design to Deployment: A Digital Architecture Strategy

The life sciences industry invests heavily in digital technologies, yet manufacturers continue to struggle with fragmented data, manual reconciliation, and limited return. This paper argues the root cause is architectural, not technological. It introduces the Intent-Driven Digital Core (IDDC) — a four-layer framework (Semantic, Event, Context, Evidence) that preserves manufacturing intent as a persistent, machine-readable construct across the full product lifecycle. Building on ISA-88, ISA-95, and ICH guidance, the IDDC makes technology transfer repeatable rather than bespoke, shifts batch disposition from retrospective reconstruction to continuous evaluation, and gives AI the contextual grounding required for trust in regulated environments.

What's inside: Manufacturing intent as the architectural anchor · the IDDC reference architecture · ontology as a contract of meaning · people as an architectural dependency · digital technology transfer as the ultimate stress test · AI enablement as a consequence of architecture · an alarm-management case study in cell therapy manufacturing.

Sandboxes, Stewards, and the Governance Gap

An Organizational Model for AI in Regulated Manufacturing

Centralized AI governance in regulated manufacturing creates a structural paradox: restrictive policies meant to manage risk drive AI use underground into a shadow tier — invisible, ungoverned, and disconnected from organizational learning. In April 2026, the FDA's first warning letter citing inappropriate AI use in pharmaceutical manufacturing turned that shadow tier from a theoretical risk into a documented enforcement trigger. This paper proposes an integrated architecture — sandbox environments for structured experimentation, a competency-tier model coupling training with tool access, and AI stewards (named individuals with dual fluency in domain expertise and AI literacy, carrying explicit GxP accountability). The components exist independently across industries; the argument is that they are interdependent by design. The integration is the contribution.

Also introduces: the Attributable–Observable–Correctable design discipline for human oversight of AI outputs, and the GxP-steward specification for functions where AI influences quality decisions.

Batch Disposition: A Shared Vision and Practical Framework for Digitized Transformation

IT Digital & Data Phorum · Co-author

Batch disposition today remains largely manual, fragmented, and inconsistent across sites — a poor fit for the complexity and pace of modern biopharmaceutical supply. This BioPhorum paper sets out a shared industry vision and a practical framework for moving from retrospective, paper-bound release toward automated, data-driven, globally harmonized disposition. It names the core challenges, the strategic opportunities, and the guiding principles teams can use to assess their maturity and chart a path forward.

AI-Ready Data: A Practical, Risk-Based Framework

IT Digital & Data Phorum · Co-author

The barrier to trustworthy AI in biopharma is rarely the model — it's whether the underlying data is ready. This BioPhorum framework defines what makes data AI-ready — suitability, observability, machine-understandability, and availability — and shows how those requirements scale with use case and risk. It confronts data fragmentation, regulatory complexity, and organizational barriers, and offers actionable practices: treating data as a product and building the knowledge layers that make AI adoption scalable and defensible across discovery, manufacturing, quality, and supply chain.