Principal search path — my private kit
This is not a product. It’s the kit I actually use: one search, one pitch, one scoreboard, planned as a nine-month arc from day one.
I am not running a “close track” and then a “north-star climb.” That split made me starve lab / FAANG warmth until I signed — which is how those loops go cold. I keep north-star threads warm in parallel the whole time. What changes is how much energy they get, not whether they exist.
Brand (one line, every call): I design and ship governed agent platforms — orchestration, governance, access-aware RAG, eval gates. Not chat demos.
Verbal heroes: AegisAI + Enterprise RAG. Lucid is production (P). The public spine is inspectable proof of the same principles (O). I never blend those.
FDE / Applied seats: same search, same kit. Discovery → wedge → HITL is a method I already practice in fde-deployment/. I don’t invent a second career brand for it.
The arc I planned in advance
| Chapter | Window | Job | North-star warmth |
|---|---|---|---|
| 1 · War room | Days 0–30 | Convert loops. Lock evidence. Don’t flop live. | 1–3 thoughtful threads / week. Homework, not spray. |
| 2 · Close | Days 31–60 | Sign without going quiet. | 2–3 warm threads stay alive. No restart spray. |
| 3 · Bridge | Months 3–6 | Earn stories on the seat. Charter. One signal artifact. | 2 intros / week. Referral-first. |
| 4 · Peer seat | Months 6–9 | Lab / FAANG AI Platform / peer Principal — or a dated upgrade. | Full loops. Same pitch. Deeper scars. |
Readiness beats the calendar. If I sign on day 18, I still don’t kill north-star warmth. If I’m still in finals on day 45, I don’t pretend I’m in chapter 3.
What I’d say if someone asked “what’s your plan?”
I’m running one search. Close a Staff+ / Principal / Founding seat with real agent, RAG, or governance ownership — and keep two or three north-star threads warm the whole time so I’m not cold-starting Anthropic, OpenAI, or FAANG AI Platform six months later. Same pitch on every call. I don’t rebuild a catalog because a reject stung.
How I want to sound
| Mid-level | What I actually say |
|---|---|
| Catalog of demos | Two systems deep + a deny path |
| “We used LangGraph / RAG / GPT” | Constraint → decision → what I’d change on their stack |
| Title as proof | A standard another team reused. HITL. An eval that failed a build. |
| Perfect confidence | Demo vs Strict. What I don’t measure yet. |
| New pitch per logo | Same brand. Map their 6–12 month problem to two or three rows. |
Pass test: if a Staff engineer could steal the answer with a find-replace of company names, I add my friction, rollback, or honesty clause.
Must-pass verbal surfaces (before day 30 — keep them warm forever)
- Pitch + call flow — 01
- Multi-tenant short + deep — 02 → 10
- Two STARs cold — EDI platform + Kafka in 03 + 12
- Coding honesty (Python/TS primary) — 11
- Spine doesn’t flop on a live review — 04
Chapter 1 — War room (days 0–30)
Do this order. Don’t skip the evidence lock to send more apps.
- Private evidence bank — fill every CONFIRM
- Resume and public footprint
- Pitch + capability map
- Multi-tenant isolation → deep path
- STAR pack + behavioral map
- Coding refresh
- Spine health
- Accept / reject rubric
- Weekly scoreboard — one board for the whole arc
Daily quota: 6–10 high-signal actions. Active loops first. Then new apps. Then north-star warmth. Don’t skip the intros because follow-ups ate the day.
Day-30 gate: ≥1 strong verbal/written offer or ≥2 finals. Pitch + multi-tenant fluent. Zero unexplained demo failures. No open LOCKED/CONFIRM contradictions on the resume. North-star threads still exist.
Scripts I actually use in chapter 1
“Walk me through your background.”
Full or 90s pitch (01), then their 6–12 month problem. I do not tour five demos.
“Why are you looking?”
I want the next seat to own agents, retrieval, and the control plane around tools that can do damage — not another chat prototype. Lucid is complete enough to defend. I’m not leaving bitter, and I’m not using anyone as a stepping stone.
“What’s your greatest strength?”
I get uncomfortable when an agent can write to a CRM or publish without anyone in the loop. So I put policy and HITL in front of the scary tools, and I want an eval that can actually fail a build. Lucid 0→1 with outcomes I can walk; one public spine a panel can poke. That’s the pair.
“What’s a weakness?”
I’m not day-to-day Go, and I’m not a daily Java agent runtime. Java and Spring are real history from Volvo-scale ecommerce. If this team is Go-primary I’ll say that up front. I won’t fake fluency.
“Design multi-tenant agents.”
Threat model first (02). Mark Demo vs Strict. Invite the deep path if they want layers.
“Tell me about a failure.”
A malformed ASN passed schema validation. SAP accepted the transaction and never created the inbound delivery. Silent. We caught it 18 hours later in reconciliation. Fix: poll SAP for inbound-delivery creation within five minutes of EDI receipt and escalate if it’s missing. Schema-valid is not business-valid. Full STAR: 03.
“Walk me through a project.”
STAR 1 — EDI platform (~60s) in 03. LangGraph routers, shared RAG/vector context, Kafka to microservices. Close on the SAP-director fight and the silent ASN, not a catalog of agents.
“Why this company / this role?”
I did the homework on [specific public signal — team, problem, not the logo]. I want ownership of agents, retrieval, or the control plane around tools that can do damage. If that’s not this seat, we should say so now.
“Are you Principal or Staff?”
Lucid titled me Sr Staff. The work sat across programs. I’m happy to be leveled here. What I won’t do is shrink it into a single-team feature lead. Full scripts: 18.
“Were you a Google employee?”
No. Apple and Google were client engagements through Sparity. I don’t claim a Google W-2 I don’t have.
“Comp expectations / when can you start?”
Numbers stay offline. On the call: I’m looking at Bay Area Sr Staff AI total-comp, not a Senior band with an AI sticker. Start date is workable against a professional Lucid notice — I won’t leave a team in a mess to look hungry. Rubric: 05.
What I refuse on a call. A catalog tour. A per-logo pitch. Invented P95s. Pretending Demo SSO is production. Calling Lucid’s public-repo cousins “what we ran in prod.”
Chapter 2 — Close without going quiet (days 31–60)
Shorter pitch. Sharper close: timeline, written offer, rubric. Same brand. 2–3 north-star threads stay warm. Spray volume is how finals die.
On the phone: What’s the decision timeline, and who owns the next step on your side? Full scripts: 07.
Chapter 3 — Earn the next story (months 3–6)
Useful on the seat by day 30 without becoming the approval bottleneck. Manager can say what AI surface I own. One recommender talks platform / RAG / governance, not iOS. One linkable deep proof. Five to ten people inside north-star orgs would take my call.
Those north-star intros started in chapter 1. I am not “opening the climb.” I’m turning warmth into referrals.
On the seat, week 1: I want the first 90 days to produce something another team can reuse — a deny path, HITL on a scary tool, or an eval that can fail a build. Full scripts: 13 · 14.
Chapter 4 — Sit as a peer (months 6–9)
Real loops past recruiter. Panels where I sound like a peer on isolation, evals, governance. Leveling story that doesn’t apologize for Lucid’s title ladder. North-star offer or a written bridge upgrade with a dated next gate — not vibes.
If they ask Staff vs Principal: Lucid titled me Sr Staff. The work sat across programs. I’m happy to be leveled here. What I won’t do is shrink it into a single-team feature lead. Full scripts: 18.
Cadence
One Friday board: Weekly scoreboard. Chapter checkbox changes. The north-star column never goes blank on purpose.
Outreach I actually send
That’s it. I don’t paste seventeen product URLs.
I will not
New repos. Catalog expansion. Per-company pitch cards. Invented SLOs. Blending open repos into Lucid production. Committing offer PDFs or exact asks into git. Starving north-star warmth “until I sign.” Saying “stepping stone” about any seat I’m on.