Behavioral calibration map

The STARs in 03 are opening moves from the master story bank — lived Lucid/Volvo/O stories, not theme-sliced slogans. Graded behavioral/ entries are longer Principal versions. All behavioral files stay practice-status: draft until I pass timed mocks with private notes.

Map (this order)

Kit STAR	Graded behavioral	Resume / bank anchor	My job
1 — EDI platform	04 Payments/EDI	20k suppliers; LangGraph router + model router; shared state + RAG/vectors; Kafka to microservices; 34 specialists; $10M + $7M net; 200→25	Don’t mash with 10→2. Don’t guess vector product / model SKUs.
2 — Kafka MSK	06 Influence without authority	$1M+; 4-week pre-prod; infra director converted	VERIFY rebalance mechanics before cooperative-sticky
3 — Lucid payments	04 (payments half)	Stripe + Klarna/Giropay/iDEAL; GIB Gulf; Chase ACH; 9+ US/India/APAC	CONFIRM retry/fallback/reconciliation
4 — Governance	07 Failed bet + 03 Security if portfolio-only	Lucid control plane; HITL 8%→2.3%; public AegisAI is O	Don’t claim GitHub binary ran inside Lucid
5 — Access-aware RAG	Design answers + eval story	Lucid P + ERAG O	Same P/O split; 3-pt / 99.6% are CONFIRM if legal flinches
Volvo 0→6	09 Mentoring	9 interns 100% convert; 20% revenue contributed	OPEN: named Sweden-HQ fight — do not invent
Staffing 10→2	01	Separate from 200→25	Don’t mix in one sentence
Peer conflict	08	SAP directors vs owned EDI	Volvo HQ still OPEN
FinOps	02	Portfolio-safe	Keep O unless Lucid FinOps is confirmed
0→1 ambiguity (O)	05 Content Factory	ACF publish scope	Not a Lucid substitute

FDE tilt: same STARs + Why FDE. Not a second career story.

Calibration protocol

As candidate

Fill CONFIRM rows in the evidence bank.
For each mapped behavioral: write 30s / 2min / 8min versions.
Blind timed mock — no notes.
Only then mentally promote off draft.

How I’ll be judged

They listen for:

Specific decision + discarded alternative
Metric with provenance (or honest “I won’t invent”)
Setback + fix (silent ASN, HITL 8%→2.3%)
Org reach beyond my team — a standard someone else reused
No blending of open demos into employer production

If I sound like a README, I fail. If I sound like an operator with scars, I pass.

How I open a STAR (so it doesn’t sound like a resume)

Here’s the decision I made, the option I discarded, and what broke when we were wrong. Then the number — and how we measured it. If I can’t name the discarded option, I’m not ready to tell this story.

I do not open with headcount, title, or “I led a team of…” unless they asked about leadership. I do not mix the public repo into the Lucid sentence.

Questions that will get asked

EDI / 200→25 — STAR 1

20k suppliers. Cheaper vendor was the tempting path. We built the layer. Onboarding 200 to 25. $10M licensing gone, $7M net. SAP directors pushed back; I kept invoice gen on SAP and shipped anyway.

10→2 staffing — behavioral/01

Different workflow from 200→25. Eight roles went away — I won’t call that redeployed if it wasn’t. Architecture was mine. Employment decisions sat with the business.

Influence without authority — STAR 2

Infra director, Confluent, four-week pre-prod, he presented the win. Spoken draft: Q9.

A bet that failed / what broke — silent ASN, then autonomy→HITL. Q7 · Q10.

Disagreement with peers — SAP directors. Volvo HQ is OPEN — I will not invent it.

Production incident — the ASN/SAP miss is the incident. Q14.

Minimum before the next onsite

STAR 1 (EDI) + STAR 2 (Kafka) fluent, including silent ASN and “what if pre-prod failed”
Dual-UI (internal vs supplier portal) said out loud twice — it’s new muscle
Kafka rebalance paragraph verified against memory or dropped
Behavioral 04 + 06 timed once each
Mentoring story uses Volvo nine interns + one private mentee detail
Do not hand the resume PDF until projected $5–7M / $8–9M is gone

STAR pack
Private evidence bank
Evidence ledger: docs/CANDIDATE_EVIDENCE_LEDGER.md