# `Vannevar Labs` — Job Pipeline Surface

> Forgejo bucket: `marcus/job-vannevar`. Maintained as part of Scott's
> Staff+ SWE job-pipeline. **Surface only** — no public posts, no
> outreach sent, until `sscoble` approves.

## 1. Why Vannevar Labs

Vannevar Labs builds defense-intelligence AI for the U.S. national-security community — Mission Agents, Search, and ATO (Authorization To Operate) software. Defense intel is the canonical "feedback loop is absent" dark-factory domain: the operator's decision is downstream of agent output and cannot audit the agent's reasoning in real time. The Speckl × Vannevar demo (access control provenance compile) demonstrates compile-time policy provenance for mission-agent access control, directly mapping Scott's OneBrief defense-tech and B.Well regulated-platform experience to Vannevar's mission-agent + ATO surface.

## 2. JD snapshot

| Field | Value |
|---|---|
| Title | Engineering Manager, DevOps / Engineering Manager, Security Platform / Senior Software Engineer - Agents Platform |
| Department | DevOps / Security Platform / Mission Agents |
| Level | EM / Senior (Staff-equivalent for EM roles) |
| Location | Remote (US) / San Diego / New York (varies by role) |
| Remote/hybrid/on-site | Remote (for most roles); on-site for some (San Diego, New York) |
| Comp range (base) | Not posted (Series B; estimated $180–200K base) |
| Comp range (total) | Estimated $250K+ all-in (equity at Series B) |
| Req ID | Via Greenhouse — re-verify at outreach time |
| Posting URL | https://www.vannevarlabs.com (careers) |
| Updated | 2026-07-07 (v3 shortlist research pass) |
| Tech stack named | K8s, CI/CD, DevOps, security platform, agentic AI, TS/SCI environment |
| Soft requirements | Defense-tech experience, TS/SCI clearance (for some roles), mission-agent development |
| Years experience | 8+ (EM-level); 5+ (Senior SWE) |

### Additional open roles (verified live on Greenhouse, 33 total)

- **Engineering Manager, DevOps** (Remote) — *direct title match*
- **Engineering Manager, Mission Agents** (New York) — agentic AI leadership
- **Engineering Manager, Security Platform** (Remote) — direct match
- **DevOps Engineer** (Remote)
- **Senior Software Engineer - Agents Platform** (Remote) — agentic AI infra
- **Senior Software Engineer - ATO** (San Diego) — Speckl adjacent
- **Senior Software Engineer, Fullstack** (Remote)
- **Senior Software Engineer, Fullstack (Mission Agents)** (Remote)
- **Senior Software Engineer, Search** (Remote)
- **Forward Deployed Engineer** (TS/SCI Required) — multiple US locations

### Full JD body

*Verbatim JD text should be pasted here from the Greenhouse job board at outreach time. The v3 shortlist captured the role titles, locations, and department context; the full JD body will be fetched from `https://www.vannevarlabs.com` or Greenhouse before outreach is approved.*

**Key verified product context from v3 research:**
- Defense-intelligence AI platform for the U.S. national-security community.
- Verified tagline: *"Restoring Deterrence, Reclaiming Advantage."*
- Products: **Mission Agents** (agentic AI for intel workflows), **Search**, and **ATO** (Authorization To Operate) software.
- Series B (~$150M+ raised; specific round size unverified in this pass).
- Headcount ~150–300 (33 open Greenhouse roles is a strong signal).
- Mission Agents = agentic AI in production where the human can't catch the bug.
- ATO = the literal compliance-automation dark-factory problem: prove your system is safe to operate against a regulator's checklist. Speckl-adjacent.
- Security Platform as a first-class team — verified open role: Engineering Manager, Security Platform (Remote).
- Defense-tech peer graph = Scott's current OneBrief network. Warmest defense-tech entry on the shortlist.

## 3. Public surface

### Hiring manager (highest confidence)

- **Name:** Brett Granberg
- **Title:** CEO / Co-founder
- **GH handle:** (to be verified)
- **LinkedIn:** (to be verified)
- **Email domain:** vannevarlabs.com (not directly verified in this pass)
- **Signals:** CEO/co-founder; publicly known from press but not directly verified in this research pass. Flagged in v3 shortlist as "publicly known from press but not directly verified."
- **Confidence:** medium (founder identity publicly known from press; direct verification pending — the v3 subagent flagged this explicitly)

### Adjacent public figures (good reply targets before DM)

- Engineering leadership for Mission Agents, Security Platform, and DevOps teams — to be surfaced via Greenhouse/GitHub/LinkedIn before outreach
- OneBrief colleagues with Vannevar connections — warm-intro path to be verified

### Highest-value reads (3-5)

- Vannevar Labs homepage — *"Restoring Deterrence, Reclaiming Advantage."* — mission framing
- Mission Agents product documentation — the agentic AI dark-factory surface
- ATO (Authorization To Operate) product — the compliance-automation dark-factory
- Open Greenhouse role descriptions for EM DevOps and EM Security Platform — the role-shape match
- Defense-tech press coverage on Vannevar's funding and product launches

## 4. Artifact plan

### Premise

The Speckl × Vannevar demo is an access control provenance compile that demonstrates compile-time policy provenance for mission-agent access control. The artifact proves Scott can build the policy-provenance layer that Vannevar's Mission Agents + ATO products need — directly mapping his OneBrief defense-tech and B.Well regulated-platform experience.

### Repo layout

```
marcus/job-vannevar/
├── README.md
└── (references https://github.com/wscoble/speckl-access-control)
```

### Artifact URL

**https://github.com/wscoble/speckl-access-control**

This is the Speckl × Vannevar demo. Build path is `git clone` + `cd speckl-access-control`. Expected deliverable is a working 4-artifact compile output.

### What it demonstrates

- Intent-level provenance compiled through Speckl for access control policies
- Content-addressed, reproducible policy artifacts for mission-agent authorization
- ATO-grade audit trail compiled by construction, not verified after the fact
- A working compile output a hiring manager can clone, build, and verify
- Direct mapping of Scott's OneBrief defense-tech + B.Well regulated-platform experience to Vannevar's product surface

### Effort estimate

<50 hours (demo already built; README + packaging for external consumption is the remaining work)

### Sketch

```bash
# Build path
git clone https://github.com/wscoble/speckl-access-control
cd speckl-access-control
# Expected: 4-artifact compile output (Speckl provenance manifests)
```

## 5. Outreach log

| date | recipient | channel | artifact attached | reply | next step |
|---|---|---|---|---|---|
| — | — | — | — | — | — |

**No outreach has been sent. No public posts have been made.** Per
directive from the parent, surface only. Outreach log stays empty until
`sscoble` approves first contact.

## 6. Sources + confidence

| Fact | Source | Confidence |
|---|---|---|
| JD body (role titles, locations) | `vannevarlabs.com` + Greenhouse (v3 research pass, 2026-07-07) | high |
| Comp range | Not posted; estimated from Series B stage | low (estimate) |
| Hiring manager name (Brett Granberg) | Public press (flagged as "publicly known but not directly verified" in v3) | medium |
| Brett Granberg = CEO/co-founder | Public press; company site JS-rendered, blocked direct scraping in v3 | medium |
| Funding (~$150M+) | Widely reported in press; specific round sizes unverified | medium |
| Artifact plan | Parent-derived (Speckl × Vannevar consulting wedge) | parent-supplied |
| Artifact URL | https://github.com/wscoble/speckl-access-control | high |

---

*Captured 2026-07-07 by the job-pipeline subagent. Parent context:
Scott's job-pipeline strategy, consult-pilot campaign. v3 shortlist
file: `~/.hermes/cache/job-search/scott-shortlist-2026-07-07-v3.md`.*