Rōvn · Pre-seed · 2026

The operating network for the healthcare workforce.

A clinician is verified once and that verified work is reusable everywhere on the network, portable and worker-owned, instead of being rebuilt from scratch at every facility. One coherent agentic system operates recruiting, credentialing, readiness, and monitoring end to end. Rōvn does the work. Source systems prove the facts. Humans make every regulated decision.

$2.25M on a $15M post-money SAFE · Atlanta, GA

The bigger vision, one system

One system, not a pile of products.

Worker-owned credential→ Operator (paid engine)→ Facility Intelligence (switching cost)→ Compounding data flywheel→ Operating network (the end state)

Order is everything: the Operator is revenue, worker-owned portability is the open corner, the data flywheel is defensibility, the network is the end state. Win the facility as a paid operator, capture the data, and density follows as a byproduct, bought with revenue, not subsidy.
Buyer one-liners. Facility: find verified workers, clear them faster, get them billable sooner. Worker: verify once, reuse everywhere.
A verification rail can be copied. The compounding outcome data, generated only by operating the workflow, cannot, the longer Rōvn runs, the further ahead it gets.
LinkedIn owns professional identity. Indeed owns job discovery. Rōvn owns the operating network for the healthcare workforce, the highest-trust, hardest-to-fake layer of them all. A system of record isn't a product. It's a category.

The problem

Facilities can't answer the one question that gates revenue.

Who can start, practice, bill, and pass audit, today?

Every facility re-verifies the same worker every 90-730 days, in parallel silos. RN time-to-fill 78 days; turnover ~$60K.
When a worker moves, the next facility starts at zero, no memory.
"Verified" is asserted without source, timestamp, or depth, now legally dangerous under Joint Commission / CMS surveyor PSV + NCQA 2024.

Why now

The regulatory window is open, and the rail is missing.

Three stacking tailwinds: NCQA's updated standards (effective July 1, 2025, 30-day ongoing monitoring), Joint Commission / CMS surveyor PSV enforcement teeth, and continued FCA exposure, billing for an improperly-credentialed provider remains exposed to False Claims Act liability under Medicare's 60-Day Rule.
CMS launched a free National Provider Directory (2026), but it's enumeration, not verified readiness. It proves the gap and that no one owns the proof layer.
AI-native credentialing is now funded and shipping (Verifiable CredAgent, Medallion CredAlliance), they target payer-side PSV, not worker→facility readiness in CAH/ASC/LTC/SNF. That's our lane.

Category & wedge

One category. One wedge.

Wedge

Receipt-bound Readiness.

A source receipt behind every green light; a named human behind every regulated decision.

The Operator workflow

One auditable pipeline.

Worker evidence→ Source receipts→ Facility rules→ Readiness engine→ Human decision→ Proof packet→ Audit trail

Readiness resolves to three answers a facility can act on: start-ready · practice-ready · billable-ready, each with the source receipt and the named human behind it.

What's defensible TODAY

The trust spine, enforced in the database, not the copy.

Live · code-enforced

Receipt-bound readiness

Migration 086 blocks any passed/cleared/ready/verified write without a valid, fresh, role/state-matched source receipt. No vendor enforces trust at the DB layer.

Live

Hash-chained audit log

Append-only, mutation-blocked at the trigger level. Replayable for Joint Commission / CMS surveyor PSV / CMS recoupment defense.

Live

Named-human decision gates

A decision can't be recorded without a named maker + a ≥40-char typed reason, the doctrine encoded as a constraint. Anti-rubber-stamp by design.

This is the asset that survives a hospital General Counsel reading every line.

How we use AI

Rōvn does the work. The human decides.

AI operates the workflow · source systems prove the facts · humans make every regulated decision.

Agents run the operational work behind four human decision gates; every output carries a depth label + source receipt.
Claude under BAA via AWS Bedrock today. We never let AI decide, credential, hire, or privilege.
Agents are a feature. The Operator is a company. We run agents internally; the external story is the operating network.

The agent stack under the Operator

Eight purpose-built agents on one spine.

applicant→ screened→ documents→ source verified→ credentialed→ ready→ deployed by the facility→ monitored

Build · gated

Recruiting

Engages and screens applicants against role requirements.

Live

Credentialing

Intake, extraction, packets, gap-chase. Human approves.

Live · rails

PSV

Primary-source verification; receipts; discrepancies.

Live

Readiness

Start-ready/practice-ready/billable-ready; blockers; what needs a human.

Live

Monitoring

Sanctions/expiration surveillance; renewals.

Build

Demand

Captures open roles, shifts, urgency, coverage gaps.

Build · gated

Deployment

Recommends workers who are ready, available, approved. Recommend-only.

Build

Facility

Each facility's own rules, patterns, memory.

The AI expansion engine, how the LLM/ML story sequences

Don't train a model first. Earn the right to one.

Phase 1 · now

Use frontier AI

Claude/GPT/Llama + Rōvn workflows + healthcare data + agent execution. live

Phase 2

Proprietary data layer

Capture credentialing, verification, decision, and outcome events as structured, trainable data.

Phase 3

Fine-tune small models

Specialized models for extraction, packet assembly, compliance checking, exception handling.

Phase 4 · the option

Workforce LLM

The option the graph buys later: graph plus facility memory plus outcomes plus model reasoning on our corpus. The graph creates the model, never the reverse. later, gated

The goal isn't "build AI." It's build the operating system that teaches AI how healthcare workforce work actually gets done.

The data we train on, why it compounds

Proprietary workforce data no competitor can export.

What we accumulate

The workforce graph: workers → credentials → facilities → readiness → decisions → outcomes.
Source receipts, requirement versions, human decisions, blocker patterns, cycle times, acceptance/rejection reasons.
Every receipt is hash-bound and audit-linked, verifiable training data, not scraped text.

Why it's defensible

❌ "No customers, building an LLM" → expensive science project.
✅ "Customers + proprietary workflow data → training specialized models" → defensible AI infrastructure.
Like Harvey · Abridge · Sierra: value from owning a workflow + data, not competing with frontier providers.

First proprietary models (after data exists)

Structured prediction, not an LLM.

Credentialing cycle-time predictor, where a file will stall, and the next-best action.
Readiness-risk predictor, who is at risk of missing clear-to-start by a date.
Proof-packet acceptance model, likelihood a facility human accepts a packet as-is.
Facility-fit / clearance-probability model, operational fit on verifiable criteria only.
Gradient-boosted trees beat LLMs on this tabular data. Ready-score calibration comes after real outcomes accrue.

How we're different

Everyone owns a slice. We connect the slices.

Player	They own	They don't own
Indeed / LinkedIn	Discovery, profiles	Verified readiness, facility rules
symplr / Verifiable	PSV workflows, provider data	Worker-owned network, facility memory, readiness
Clipboard / ShiftMed	Labor deployment	Deep credentialing, readiness, facility memory
ATS / hospital HR	Pipelines, records	Source-verified readiness across facilities

Rōvn's wedge: verified readiness + facility intelligence + worker-owned evidence. We are not a staffing agency, job board, or scheduler, no placement or commission fees, ever.

What we are not

Rōvn is not a staffing company. That's the point.

We are NOT

A staffing agency · a job board · a per-diem marketplace.
A scheduler · a recruiter-only tool · "Indeed for healthcare."
An LLM company chasing a foundation model.
No placement, commission, or success fees. Ever.

We ARE

The operating network for the healthcare workforce, Rōvn does the work, the facility decides.
The verified-readiness layer underneath every hire, reusable everywhere.
A worker-owned evidence network compounding into the system of record.

Staffing marketplaces are crowded, low-margin, and copyable. The trust layer is none of those, and no one owns it yet.

What we're adding

Five additions that turn the spine into a product.

Customer-facing

Proof Packet Export

Downloadable packet: requirement version, source-receipt appendix, human-decision appendix, freshness clock, packet hash, acceptance status.

Outcome capture

Facility Acceptance Outcomes

Accepted / rejected / returned / delayed + structured reason code + time-to-start/practice/bill. The first real outcome labels.

Honesty

Source Adapter Readiness

Live / partial / manual / planned by role, source, jurisdiction. Kills the "36 live sources" overclaim.

Worker-owned

Passport Consent UX

Share, scope, expire, revoke, and see who accessed evidence. The consent layer that makes reuse legal.

Pilot deliverable

Design-Partner Closeout

Roster imported, receipts created, blockers resolved, packets produced, cycle-time impact. The proof artifact.

Build-toward, future compounding layers

The expansion, framed honestly.

Cross-facility evidence reuse

Facility A accepts → worker consents → Facility B consumes the Passport → Rōvn maps gaps → the worker clears faster. The network unlock.

Facility Memory

Learns requirement versions, decisions, blockers, cycle times, acceptance reasons, the memory layer for workforce ops.

Readiness-aware deployment

We recommend workers who are ready, available, and approved. Readiness-aware recommendations, never auto-placement. The facility decides and hires directly; no placement fees, ever.

Readiness-aware scheduling

Later only: who is available, credentialed, compliant, facility-cleared, monitored, and audit-ready. Recommendations, not a scheduler. We don't lead with scheduling.

Honest current state

What's built vs build-toward.

Capability	Status
Receipt-bound readiness spine (086) · hash-chained audit · human gates	Built
Readiness model (clear-to-start/practice/bill) · worker trust record · PSV rails	Partial
Facility Intelligence (configurable rules) · ai_runs ledger	Partial
Cross-facility evidence reuse · Facility Memory (learning) · proprietary models	Build-toward
Staffing / placement fees · auto-scheduling · finished proprietary LLM · paying customers	Not claimed

Pre-launch by design. Zero invented traction. The round converts proof architecture into customer proof.

Team, built to execute procurement-grade healthcare infra

A team that has already shipped and sold healthcare software.

Founder & CEO

Giles-Evan Mboumi

Designed and deployed Rōvn's production verification architecture. 3 yrs healthcare commercial sales (Boehringer Ingelheim, Cardiovascular-Renal). Built BOVYN, a real-time futures system he coded himself, and a 200+ customer business from zero. Led operations on government contracts across France, Gabon, and the U.S. Owns vision, capital, GTM, pilots, recruiting.

Co-Founder & COO

Christian Montgomery

Built Rōvn's deployed attestation pipeline as founding engineer. B.S. Cybersecurity (Kennesaw State, 3.96 GPA); ran AD / M365 / network security for 200+ users. Now owns operations, finance, delivery, security & infra, and the technical-to-non-technical liaison. Security depth foundational for a HIPAA-aligned product.

Chief Technology Officer

Abhishek Jha

~6 yrs SWE (Accenture; Senior SWE at Eruvaka), distributed systems, AWS, agentic AI; MS CS (Texas A&M, Corpus Christi). At Braintree Health, rebuilt a healthcare data pipeline from over a week to ~90 min across 15M+ records. Shipped 50+ products. Owns architecture, implementation, infra, reliability.

Chief Product Officer

Gokul Shanmugam

Engineer at Aetna (payer-side), then 3.5 yrs building EHR software at Athenahealth (Software Engineer II; log4j Hall of Fame; hackathon runner-up), plus RCM automation at Commure. The deepest healthcare-software résumé on the team. Owns product, roadmap, verification-engine behavior.

Clinical Advisor

Dr. Danielle K. Miller, DNP RN

Former CNO; ex-Amazon Principal PM (Health Equity) + AWS Healthcare Exec Advisor; former Huron Sr Director; 20+ yrs; founder, The Pivot Nurse. She's been the buyer. Owns clinical & regulatory acceptance criteria.

Advisor

Dr. Mohammed Quadri, MD MBA

VP of Strategy, Hackensack Meridian Health (19+ yrs); active healthcare-AI investor and mentor. Brings health-system strategy depth and adoption credibility.

Strategic & GTM Advisor

Aki Hashmi

CEO, SkinSAFE & DeepStart; repeat health-tech CEO. Brings a network of providers + VCs; guides pricing + provider-group GTM. (SkinSAFE's Mayo data credential, not a Rōvn, Mayo partnership.)

Chassis & counsel

engineering partner under NDA + Acevedo

engineering partner under NDA, 10-yr healthcare-SaaS firm, 50+ live products, 0 HIPAA violations: enterprise capacity without the cost. Counsel: Jason T. Acevedo (MSA/DPA/BAA/SAFE/IP).

The flywheel (build-toward)

Every workflow makes the next one cheaper.

More workers→More verified passports→More facilities→More facility memory→Faster starts→More outcome data→Smarter models

The unlock: cross-facility Passport consumption, Facility B reads Facility A's verified evidence with worker consent.
Honest today: this is build-toward. The round funds the first turns of the wheel with design partners.

Go-to-market & 90-day proof

Free 90-day design-partner Readiness program.

GTM

Atlanta-area facilities + provider groups (shorter sales cycles).
Free 90-day Readiness pilot in exchange for feedback + data access.
Closeout report = the conversion artifact.

The next milestone (kept simple)

One real roster imported.
One real live source receipt on a consented worker.
One human-approved proof packet.
One measured operational improvement.

The raise

$2.25M on a $15M post-money SAFE.

Structure

$1.8M minimum viable · $2.25M target · $2.75M hard cap.
$12M, $15M post-money range.

Use of funds

Design-partner onboarding + live source verification.
Readiness workflow completion + proof packets.
Product hardening + compliance/security (incl. tenant-isolation hardening).
GTM conversion.

Round thesis: Rōvn has built the trust spine. The round converts proof architecture into customer proof.

What this round buys

From trust spine to live network.

$2.25M turns proof architecture into customer proof, and lights the first turns of the flywheel.

Quarter 1-2

Design partners live

3-5 facilities on the free 90-day Readiness program. First real rosters, first live source receipts.

Quarter 2-3

Proof packets + outcomes

Human-approved proof packets shipped; acceptance outcomes captured, the first proprietary data.

Quarter 3-4

Evidence reuse turns on

Facility B consumes Facility A's verified Passport. The network starts compounding.

Into Series A

Models + paid motion

First prediction models on real outcomes; pilots convert to paid. The system of record begins.

The next milestone, kept simple: one real roster · one real source receipt · one human-approved proof packet · one measured improvement.

Rōvn

The operating network for the healthcare workforce.

One coherent agentic system operates the workflow; the facility makes every regulated decision. Building the network now: one real roster · one real source receipt · one human-approved proof packet · one measured improvement. Then the data compounds.

$2.25M · $15M post · Atlanta, GA · HIPAA-aligned · BAA available