AI-Native Engineering · 2025 – Present
A simple solution to a large problem
CSC was going live. At the same moment, DALA needed the same application rebuilt with different business logic, a different data schema, and different workflows. One month. Two developers. Hundreds of interdependent objects. The real problem wasn't how to build the app — it was how to design a system that builds the app automatically.
The Pipeline
3-stage agentic architecture
Each Appian object — processes, interfaces, records, data structures — was structurally identical between departments. Only the identifiers changed. That's not a development problem. That's a transformation problem. Claude Code does transformation at scale.
01 — Transform
Transformer
Ingests source XML from CSC. Substitutes all department-specific references — UUIDs, group names, folder structures, data schemas — using dala_mapping.json config
02 — Validate
Validator
Runs 12 automated checks: no remaining CSC references, UUID integrity verified, dependency tree resolved, schema correctness confirmed
03 — Package
Packager
Zips validated output into a Designer-importable package ordered by dependency depth — ready to deploy with no manual intervention
Real prompt — standard transform
Transform LCMS_CSC_createCase from reference/source-exports/
using config/dala_mapping.json.
Identify all dependencies: sub-processes, expression rules,
constants, data types, record types, groups, folders.
Recurse into each sub-process dependency tree depth-first.
Scan reference/dala-exports/ for DALA equivalents.
Add confirmed mappings to config. Flag missing precedents.
Output: dependency tree · mappings · missing precedents
· proposed clone order · ambiguous names for confirmation
Clone missing precedents CSC → DALA recursively.
Rename to LCMS_DALA_createCase, update internal refs.
Validate checks 1–12. Package as DALA_Batch_[N].