Investors
Most industrial AI projects fail because nobody can verify them. We built the one you can.
Verifiable industrial AI: signed public benchmarks, a model that can never get worse, on‑premise learning, any industry in hours — and an audit any buyer can run themselves today.
Industrial AI you can check
Ryedore predicts equipment failure, explains the cause chain, recommends the action and proves every prediction — on the customer’s hardware, with a signed benchmark and a reasoning trace behind every claim. It is the intelligence layer on top of the monitoring a plant already owns, not another monitoring platform.
On top of the monitoring a plant already owns
Historians, SCADA, CMMS and sensor gateways stay. Ryedore reads them where they are and adds the answers — remaining life, cause, action, what‑if — with a trace behind each one.
How it works →A large market with a trust problem
The failure mode is not model accuracy — it is trust: black‑box scores nobody on the plant floor can verify, models that quietly degrade, and cloud architectures that cannot hold plant data.
Industry figures; sources named on request and in the memorandum. Downtime cost is an industry estimate — never phrased as “prevented by Ryedore”.
Verifiable is the moat
Four provable differences — each with the page where you check it.
swipe → to see the whole diagram
- moat 1
Signed verifiability
Results on public datasets, cryptographically signed, re‑runnable — including where we lose. Nobody in the category publishes reproducible benchmarks.
Open the signed report → - moat 2
The never‑degrade guarantee
A new model version must beat the current one on every task or it is blocked. Static tools don’t retrain; retraining tools don’t gate regressions.
See the gate log → - moat 3
On‑prem + privacy‑preserving learning
Model, data and learning stay inside the customer’s boundary — and it still learns from their labels. Cloud single‑pane platforms need the data in their cloud.
Security & certification status → - moat 4
Any industry, in hours
One model learns the failure pattern, not the sector; a new vertical starts with everything the others taught it. Vertical vendors and OEM suites are locked to their equipment.
The industries page →
Sovereignty is a feature, not a setting
Model, data, predictions and the learning loop stay inside the customer’s boundary; air‑gap capable. That is why a refinery, a mine or a hospital can run the whole platform — and why cloud single‑pane platforms cannot follow.
The picture and the certification register →
A lab and a platform — ten programs, each with a signed proof point
Built on our own industrial multi‑task encoder — a shared cross‑industry representation, no third‑party model anywhere in the stack — with on‑site sovereignty and a self‑improvement loop that already fires. Each program extends a module that is measured today; each lives on the page that owns it. roadmap · not shipped
- 01Our own industrial multi‑task encoder, at scaleHow it works →
- 02Privacy‑preserving learning across sites — the collective immune systemSecurity →
- 03Certified serving — promotion attestationsVerify →
- 04Differentiable physics twinsWhat‑if Lab →
- 05Experiments as interventions — causal discovery at scaleSpecialists Lab →
- 06A sovereign reasoning model for RyekronixRyekronix →
- 07Edge inference next to the PLCSolutions →
- 08The label flywheel, acceleratedIndustries →
- 09Self‑improvement as a research engineHow it works →
- 10Verification as a productVerify →
This is why engineering & AI research is 32 % of the round and the GPU fleet sits inside infrastructure: the advances are compute‑ and research‑heavy, and each ends in a signed proof point — the standard this company holds itself to.
Signed on public data — wins and losses
The headline results below are signed and re‑runnable; the report lists where we still lose.
On NASA’s multi-condition turbofan data (FD002) our remaining-life model reached 17.85 RMSE — better than the published Li 2018 (22.36) and Chen 2020 (21.24) methods.
With real failure labels, our anomaly detector reached 0.992 AUC on FD003 versus 0.961 for the standard method.
Our production forecaster’s 90% interval actually covered 91% of outcomes on multi-condition sensor data (0.303 RMSE) — calibrated, not just accurate.
Production‑ready, pre‑revenue, founder‑funded — raising one $400M growth round
A single commitment, drawn in three milestone‑gated tranches ($150M · $125M · $125M): at close, at ≥ $25M recurring ARR, at ≥ $100M recurring ARR. One negotiation; capital to IPO‑readiness; the discipline of gates. target
- Why one round
Certainty of capital to IPO‑readiness and one set of terms; management operates instead of fundraising through two more market windows.
- Why this size
A lab‑plus‑platform capital profile: an own industrial multi‑task encoder, a GPU research fleet and ten research programs — the use of funds is itemised, and engineering & AI research is 32 % of it.
- Why it is disciplined
Only the first tranche is drawn at close; the rest is released on recurring‑ARR and reference milestones, never on projections. A missed gate defers the tranche.
- Pilot & Small Site tiers GA
- signed benchmark bundle published
- Standard GA
- first paid deployments
Funds: Standard → Enterprise GA build‑out · North‑America enterprise team · GPU training fleet · SOC 2 / ISO 27001 · first 25–40 customers
- Professional & Enterprise in production
- NRR ≥ 120 %
- Strategic GA
Funds: EU / APAC / ME go‑to‑market · OEM / integrator program (Flagship tier) · compliance packs at scale
- international revenue ≥ 20 %
- EBITDA break‑even in sight
Funds: Category leadership · selective acquisitions · IPO readiness · strategic reserve
One commitment, one cap‑table event. A missed gate defers the tranche — it is never drawn on projections.
Use of funds — whole round
- Engineering & AI researchOwn industrial multi‑task encoder at scale, federated site‑to‑centre learning, differentiable physics twins, causal discovery, a sovereign reasoning model for Ryekronix, certified serving32 %
- Sales & go‑to‑market (global)North‑America enterprise team from tranche 1; EU / APAC / ME from tranche 2; vertical AEs; OEM / integrator channel23 %
- Infrastructure & operationsGPU training fleet, on‑prem deployment tooling, 24/7 follow‑the‑sun support15 %
- International expansionRegional entities, local compliance, in‑region solution architects10 %
- Selective M&A / strategic reservePoint‑solution tuck‑ins, strategic partnerships — board‑approved, tranche 310 %
- Compliance & certificationsSOC 2 Type II, ISO 27001, FDA / NERC / OSHA / ISO 55000 packs5 %
- Working capital & reserveOperating reserve between tranche gates and spend5 %
Business model & unit‑economics targets
Seven published tiers from the paid entry tier (40 assets) to Strategic; six pricing components including an annual license fee; three‑year terms. Land with the entry tier, expand by tier as the flywheel produces labelled failures. Pricing is public →
Targets are targets — labelled as such everywhere on this site. Team, capital deployed to date and deployment count are stated in the confidential memorandum, not on this page.
One number set. Labelled. Checkable.
Every number on this page has a status — signed, industry figure, or target — and a source. The confidential memorandum has the tier‑driven financial plan and the same discipline.
What we will not claim
- No revenue. We are pre‑revenue and say so above; nothing on this page implies otherwise.
- No customer names. Deployment work is stated in the confidential memorandum; we put no name on this site until the customer has agreed to be attributable.
- No across‑the‑board superiority. The signed report lists the benchmarks where we lose alongside the ones where we win — the losses are part of the download, not a footnote.
- No certification badges we don’t hold. The register states each framework’s real status — aligned, in progress or planned.
- No targets dressed as results. Every forward‑looking figure carries a “target” label here and in the memorandum.
The same discipline that produces this list produces the numbers you can check: verify them before the first call →
Seven answers
- Is this real or a concept?
- Real and running. The benchmarks are signed and re‑runnable on public data (see Verify), the platform serves predictions on‑premises in under ten milliseconds, and any buyer can run the audit on their own data today.
- What does Ryedore actually do?
- It is verifiable industrial AI: it predicts equipment failure with a remaining‑life range, explains the cause chain, recommends the action, and proves every prediction with a trace — on the customer’s hardware. Two agentic systems sit on top: Ryekronix (acts, with human approval) and the Specialists Lab (experiments, never touches the plant).
- Who is the customer?
- Asset‑heavy industrials — manufacturing, oil & gas, utilities, mining, chemicals, pharma and more — starting with the mid‑market that Tier‑1 enterprise AI platforms decline to serve, then head‑on above it on sovereignty and verifiability.
- How do you make money?
- Seven published tiers — from the paid entry tier (named Pilot, up to 40 assets) to Strategic — six pricing components including an annual license fee, three‑year terms. No pilots or free trials; pricing is public.
- What is the raise?
- One $400M growth round, closed as a single commitment and drawn in three tranches — $150M at close, $125M at ≥ $25M recurring ARR, $125M at ≥ $100M recurring ARR. One negotiation, one cap‑table event, capital to IPO‑readiness, with the discipline of milestone gates.
- What is not proven yet?
- Revenue — the company is pre‑revenue and says so. Third‑party certification of the serving stack is a roadmap program, not a badge. Learning across customer sites without moving data is built and simulated but not yet run between two real sites. Customer outcomes are worked scenarios until customers attribute them. Everything else on this page is signed, attested or labelled.
- How do we diligence this?
- Start without us: run the signed audit bundle, read the never‑degrade gate log, compare the certification register against your checklist, and price a deployment from the public list. Then request the memorandum — it carries the tier‑driven financial plan under the same labelling discipline as this page.
Confidential Briefings
We are highly selective about the partners we engage. If your thesis aligns with adaptive industrial AI deployed across regulated, safety-critical industries, the founding team is open to a confidential conversation.
Direct line for investors. Confidential. Founder responds within 24 hours.
Investor Contact Form
Tell us a bit about yourself and how we can connect.
Direct Contact
investors@ryedore.comHeadquarters
Ryedore Inc.
Suite 217, 2171 South Trenton Avenue
Denver, Colorado 80231
Confidentiality
All investor discussions are treated as strictly confidential. NDAs available upon request.
Request the memorandum.
One email: investors@ryedore.com — or use the form above. We’ll send the memorandum and the signed benchmark bundle.