Define and evolve the canonical data model (Person, Employment, benefit domains), the Global ID scheme, and the custom-attribute extensibility model
Design the canonical Core API resource catalogue — contract-first OpenAPI v3, shaped by the canonical model rather than a 1:1 port of 300+ legacy APIs — plus BFF screen-composition conventions
Build the serving layer: BFF, API gateway, and process services as the sole consumer boundary, with CQRS read APIs and a bounded consistency window
Build the mediated write path (UI → ODL → system of record): sync-confirm by default, async-accept writeback via saga, transactional outbox, and compensation, with read-your-writes and pending-write handling
Build the eligibility/personalization rules layer: precompute parameters and evaluate expressions against stored data, reusing an existing evaluation library
Build the practitioner console: crosswalk editor, reference-data management, survivorship and match-threshold editors, and a daily-use steward queue — with roles, permissions, and audit
