Private evidence bank

Top-performer rule: every metric in a STAR has a provenance sentence. This page drafts answers from my resume and public footprint.

• LOCKED — safe to say from resume/portfolio as written. Still don’t invent employer adoption of public repos.
• CONFIRM — I must edit before a live panel. Private finance / people / SLO detail.

Do not paste confidential customer names or unreleased Lucid docs here if this repo is shared. Prefer a private Notes copy for CONFIRM fills.

Identity / footprint (fix these first)

Ledger Q&A — spoken drafts

Q1 — EDI savings ($10M licensing + $7M net)

LOCKED 2026-08-13 from master STAR bank (candidate-confirmed realized — not a projection).

At Lucid I led architecture for an in-house EDI/supplier platform that replaced TrueCommerce. 20,000+ suppliers, native and non-native. Two numbers, don’t add them into one vanity total: $10M TrueCommerce licensing eliminated, and $7M annual net after platform operating costs. Kafka/MSK is a separate $1M+ infra save.

Resume PDF still stale: “projected $5–7M” and/or “$8–9M” / rolled-up “$10M+”. Do not hand that PDF. Spoken uses this bank. Finance-line baseline/window still CONFIRM if they want an audit trail.

Fallback if they only have the old resume in hand:

The PDF undersells it. Realized, not projected: ten million licensing off TrueCommerce, seven million net. Happy to walk architecture and adoption. Finance-line provenance I can confirm with recruiting.

Q2 — Kafka/MSK $1M+

We migrated from Confluent to Kafka/MSK with full platform ownership transfer. Resume outcome is $1M+ annual infrastructure savings. Zero cutover incidents. Infra director presented the results after sitting in a four-week pre-prod.

CONFIRM: annualization method, services included, before/after bill. VERIFY before rehearsing the cooperative-sticky rebalance paragraph — category is partition rebalancing; mechanics must match what I actually changed. Story: STAR 2.

Fallback if unconfirmed: “On the order of a million-plus a year in infra after we took Kafka in-house. I can walk the ownership transfer. The finance-line math I’m happy to confirm offline.”

Q3 — Staffing: 200→25 (onboarding) vs 10→2 (other workflow)

Lead with 200→25 — onboarding team on the EDI platform (STAR 1). HITL only on true exceptions.

Supplier onboarding went from 200 people to 25 because the orchestrator takes native and non-native suppliers from first contact to go-live. HITL only on true exceptions.

10→2 is a different July confirmation about a supply-chain workflow — eight roles eliminated, not redeploy theater. Do not mash 200→25 and 10→2 in one sentence.

CONFIRM on 10→2: decision role (recommend vs approve), period, volume/quality, support for affected people.

People-care line: “I recommended the architecture. Employment decisions sat with the business. I won’t pretend I signed the paperwork if I didn’t — and I won’t soften eliminations into redeploy theater.”

Q4 — Lucid production vs open proof

What I say: “Same architectural principles. Lucid is P. The public spine is O.”

Q5 — Volumes / latency / quality

CONFIRM only — do not invent P95.

I won’t invent P95s. What I will stand on: 20k suppliers, 34 specialists, three orchestrators, the silent-ASN incident and the five-minute SAP confirm, HITL 8%→2.3% on new non-native types. If they want a latency table, that’s CONFIRM or I show the kind of adversarial RAG suite in the open — O, not a Lucid SLO slide.

Q6 — Personal vs team ownership

I owned architecture, ADRs, and cross-program technical direction across six programs, leading 8–10 senior engineers at Principal scope. Implementation was a team sport — I say “I” for decisions I drove and “we” for delivery.

CONFIRM: one example decision I alone made vs co-owned.

Q7 — Hardest failure / rollback

Primary (LOCKED from STAR bank): silent SAP inbound after a schema-valid ASN. STAR 1 follow-up.

Schema validation passed. SAP accepted the transaction. No inbound delivery. Eighteen hours of silent wrong. We poll for creation within five minutes now. I will not invent a dollar cost for that window.

Secondary: autonomy-first EDI agent reversed before it became default — Q10. Deny-path evals late is a process miss, not a substitute SEV.

Do not invent a third incident.

Q8 — Mentee example

At Volvo I mentored nine interns, 100% converted to full-time.

CONFIRM: one named-private mentee: start → intervention → independent ownership → outcome.

Q9 — Influence without authority

Scaffold: behavioral/06 + STAR 2.

Infra director: production is stable, don’t touch Confluent. Legitimate. I didn’t win a meeting — four-week matched pre-prod, SMT rebuilt as Lambda, dual-write with lag monitoring. He sat in pre-prod and presented the win. $1M+ annual. Influence was uneven on residual paths — I won’t claim everyone converted.

CONFIRM: which clusters moved first, residual holdouts, annualization math. VERIFY rebalance mechanics before saying cooperative-sticky.

Follow-up — “what if a team still refused?”

Stay on Confluent that quarter if pre-prod fails. Narrow the next ask to a greenfield producer. A political win that reverts Monday isn’t influence.

Q10 — Failed bet

Two I can walk — pick one per panel, not both as humblebrags.

1. Autonomy-first EDI agent → specialists + HITL before it became the production default. behavioral/07.
2. Schema-valid = business-valid — the ASN/SAP silent accept. That’s the scar I lead with if they ask “what broke.”

I won’t invent an outage dollar. What stuck is a promotion gate for autonomy, and a post-import confirm — not a feature flag someone turns off under deadline.

Q11 — Peer disagreement

Lead: SAP directors vs owned EDI platform (STAR 1). Delivery-speed vs ownership is the same family as behavioral/08.

SAP directors pushed vendor relationships and reliability fear at 20k-supplier scale. I spent a month on architecture + POCs, got VP sign-off. They still didn’t help in implementation. I conceded a pure SAP-exit — kept invoice generation on SAP — and refused “forever TrueCommerce.” Names stay private.

OPEN: Volvo Sweden-HQ conflict — I do not invent one. Honest line in STAR pack — Volvo.

Follow-up — “what if they still refused?”

Escalate on risk and ownership, not personality. Disagree-and-commit if leadership accepts residual vendor lock-in. I won’t half-implement to prove them wrong.

Q12 — Setback / rollback (same family as Q7)

Use the silent ASN or Kafka “what if pre-prod failed.” Do not invent a second SEV.

Q13 — “Tell me about a time you were wrong”

Silent ASN (schema-valid ≠ posted). Or HITL 8% misclassification on new non-native suppliers — we used the queue as labels and cut to 2.3%. One story, not three humblebrags.

Q14 — Production incident I personally led

The ASN/SAP miss is the incident until a dated SEV with a dollar cost is CONFIRM-filled. Kafka cutover risk window is the other speakable path (zero incidents, but the dual-write was the danger).

Q15 — Diploma + start date

CONFIRM diploma string + first professional start month → lock years count.

If they challenge 19 years vs 2009 on the PDF:

The count includes pre-2009 internships. I’ll get those rows on the PDF. Until then I can name dates and employers — I’m not going to let the math look like a trick.

I / we cheat sheet

Related

• Resume and public footprint
• Behavioral calibration
• Org ledger: docs/CANDIDATE_EVIDENCE_LEDGER.md