AI-Driven ExtJS To React Migration For Pharma | Case Study

Case Study Details:

Industry: Pharmaceutical & Clinical Trial
Region: Global
Technology: Agent Skills · AI Multi‑Agent Orchestration · React 18 · ExtJS 7.6 · AG Grid · TanStack Query · WebSockets (STOMP/SockJS) · TypeScript Strict Mode

The Challenge

For a mission‑critical pharmaceutical BPM platform modernizing two large, production‑active ExtJS 7.6 applications became an urgent necessity.

 

The platform supported clinical trial approvals, patient data reporting, and regulatory submissions, where every action required immutable audit trails, strict PHI protection, and regulatory‑grade compliance guarantees.

 

However, the modernization program faced several high‑risk and systemic obstacles:

 

  • 152 ExtJS source files across 87,000+ lines of legacy code, built over years using complex MVVM patterns.
  • Each module consisted of multiple numbered MVVM variants, where only the highest version was canonical—requiring precise analysis before any migration.
  • Zero downtime requirement: ExtJS and React had to coexist in production using the Strangler Fig pattern, with route‑by‑route cutover.
  • Pharmaceutical compliance constraints:
    • Immutable, append‑only audit trails
    • Double‑confirmation workflows for recalls
    • Confirm‑before‑send email enforcement
    • Absolute prohibition of PHI exposure in logs, URLs, or errors
  • Real‑time BPM task queues using STOMP over SockJS demanded careful lifecycle management to avoid connection leaks and frozen approval queues.
  • Dynamic, runtime‑driven forms stored in MongoDB required schema generation at runtime, not static UI rewrites.

The complexity, compliance sensitivity, and production criticality of the system made traditional UI migration approaches unsuitable.

A compliance‑first, parallelized, AI‑driven modernization model was required.

Evoke’s Approach

Evoke Technologies designed and deployed a 12‑agent autonomous AI migration system powered by Anthropic Claude, purpose‑built for enterprise‑grade ExtJS‑to‑React modernization under pharmaceutical compliance constraints.

 

Multi‑Agent ExtJS Analysis & React Transformation

Migration was executed through a 4‑tier agent architecture, where each AI agent owned a clearly defined, non‑overlapping responsibility:

 

  • ExtJS MVVM deep analysis across all numbered variants
  • Business‑rule extraction and API contract reconstruction
  • React 18 + TypeScript strict‑mode code generation
  • State migration from ExtJS Stores to TanStack Query + Zustand
  • AG Grid server‑side pagination, bulk actions, and immutable audit grids
  • Runtime Zod schema generation for MongoDB‑driven dynamic forms

This ensured semantic‑level migration, not just UI translation.
 
Parallel Wave Execution Model

Each sprint was decomposed into six coordinated execution waves:

Parallel-Wave-Execution-Model

This parallelization compressed timelines that would otherwise take years using sequential developer handoffs.

Compliance‑First Domain Engineering

Compliance was embedded before any code was written, not reviewed afterward:

 

  • Dedicated domain agents enforced:
    • PHI handling rules
    • BPM task lifecycle integrity
    • Role‑based action rendering
    • PDR recall double‑confirmation with mandatory reason fields
  • Accessibility (WCAG 2.1 AA) was built‑in, not retrofitted.
  • TypeScript strict mode enforced system‑wide — zero any usage permitted.

Continuous Learning Through Skill Libraries

The system improved with every migrated module:

 

  • 40+ persistent skill libraries captured:
    • ExtJS MVVM patterns
    • React migration anti‑patterns
    • Pharmaceutical compliance rules
    • AG Grid server‑side best practices
    • BPM workflow state machines
  • JSON migration specs ensured full traceability from legacy ExtJS source to React output.
  • Knowledge that once lived only with senior developers became institutional, reusable assets.

The Outcomes

Metric Before AI Modernization After AI Modernization
Migration throughput Sequential, manual rewrites 6–8× parallel agent execution per sprint
Compliance assurance Manual reviews, error-prone Systematic, agent-enforced pre-conditions
UI state coverage Often incomplete 100% enforcement of all 4 UI states
Stack consistency Drift-prone at scale Locked, validator-enforced stack
Test coverage Variable, often skipped 80%+ unit + full E2E mandated
Institutional knowledge Siloed with individuals Codified in 40+ skill libraries
Production disruption High migration risk Zero downtime Strangler Fig cutover

The organization now operates a repeatable, scalable, AI‑first modernization pipeline capable of transforming regulated enterprise applications without compromising compliance, performance, or production stability.

Read the Full Case Study

Scroll to Top