Biopharma Strategy Platform

Problem

Brand strategy in pharma is not one document. It is a graph of artefacts that teams finalise, reuse, and revise for years, and the two operations that matter most are the two that are easiest to get wrong: starting a new strategy from a finished one, and removing something other artefacts quietly depend on.

Copying by hand drifts the moment the schema moves. Deleting without knowing the dependents loses work that a regulated client cannot lose.

Constraints

The platform is ISO 27001 certified and in private beta with enterprise pharma clients, so destructive actions have to be reversible and admin and user actions have to be attributable. Three core engineers cover the whole product, which rules out subsystems that need a specialist to operate.

Decisions

• The clone walks Prisma's DMMF, the generated data-model metadata, instead of hand-written per-model code.
• The archive cascades across the dependents so a restore brings the whole set back together.

Architecture

Clone engine: driven by the schema's metadata, a new field or relation is cloned because the schema says so, not because someone remembered to add it.

Dependency subsystem: a registry describes which entities depend on which, exposed through its own GraphQL contract and resolver and rendered by a registry-driven modal. Finalised entities with dependents warn before an archive or a delete.

Across the product: real-time collaboration over GraphQL subscriptions, OpenAI-backed generation features, and platform security work on OWASP/IDOR exposure and JWT sessions.

Outcome

• Two subsystems originated from scratch and in the product: the clone engine and the Edge archive/restore isolation system.
• Destructive actions on finalised work are warned about first, then cascade and reverse as a set instead of one entity at a time.
• Cloning survives schema change: the model metadata drives the copy, so new fields travel without new code.