Dual track
FDE Field Method
One method for Principal Architect and Forward Deployed loops: clarify the workflow, ship a thin production wedge under real constraints, govern irreversible actions, then hand off ownership.
This is not a sixth mega-platform — it is the operating method that turns the existing 5-spine stack into customer-ready delivery proof.
Method
Discover → score → ship → govern → eval → handoff
01
Discover
Week-1 checklist: jobs to be done, systems of record, SSO/ACLs, data reality, irreversible actions, change windows, success metric, exit owner.
02
Score the wedge
Rank use cases on value × feasibility × risk × reuse × named sponsor. Prefer ≤90-day measurable outcomes over platform theater.
03
Ship the walking skeleton
Read/assist path first. Access-before-ranking retrieval. Integrate into their identity and event planes — not a demo VPC forever.
04
Govern irreversible actions
HITL and policy gates on writes that cannot be undone. Decline-on-low-confidence beats fake autonomy.
05
Eval before rollout
Groundedness, tool success, HITL reject rate, and a baseline window for any $ savings claim.
06
Handoff ownership
Runbooks, prompt/eval owners, and a team that can operate after the FDE leaves the critical path.
Week-1 questions
What I ask before proposing architecture
- Who waits on what decision today — and what metric moves in 90 days?
- Which system is source of truth for writes?
- How does identity flow (SSO, groups, row/object ACLs) into retrieval?
- Which actions are irreversible on day one?
- What is the change-control window and who can approve production?
- Who owns the system when I leave the critical path?
Proof layers
Same spine, FDE framing
AegisAI — gateway + HITL
Policy before side effects — the FDE governance boundary.
Enterprise RAG — access-aware
Access-before-ranking for customer data constraints.
VAP — orchestration
Multi-agent workflows as a deployable wedge, not a chat toy.
Interview playbook — FDE pack
Why FDE, decomposition, discovery→MVP, client simulation.
Honesty
Evidence rules panels can trust
- Lucid Supply Chain / Commerce work is framed as an internal customer embed — not invented third-party logos.
- Open reference repos (AegisAI, VAP, Enterprise RAG, Content Factory) show how I build; they are not claims that Lucid runs those exact repos.
- Apple/Google work was client engagement via Sparity — not direct Google employment.