699 rejections. Then a new kind of hiring infrastructure.
Our founder applied to 699 jobs and was rejected by every one — filtered by keyword matchers, discarded by screening tools that had no access to the judgment of the people doing the hiring. That asymmetry is the reason Nodes exists. 13 AI agents, one intelligence layer, deployed inside your VPC or on-prem.
Three things. Held to the same standard.
Not a tool, not a workflow layer, not an overlay on your ATS. An intelligence layer that ships into your cluster, learns on your own production outcomes, and returns signed, reproducible decisions.
Built for the buyers whose security review actually reads the artifact pack.
13 agents behind one interface.
Interview, reference, writeup, calibration — coordinated by a shared state, not bolted together with API glue. Every decision carries a trace back to the agent, the weights, and the data it saw.
Single-tenant VPC or on-prem.
The infrastructure ships into your cluster — cloud VPC or on-prem datacenter. Customer-owned keys. Your data never leaves your perimeter. Revoke the key, inference stops — it's a contract-bound kill-switch, not a support ticket.
Regulated-first. Signed under NDA.
SOC 2 Type I and II held. Decision traces reproducible seven years out. Every admin, scoring, and retrain event hashed and streamed to your SIEM. The paperwork is the product.
One deployment. Ten months in production. Expanded ahead of schedule.
"Absolutely love what you guys are doing. Sometimes I sit down and start thinking of so many more things we can do with this."
What we will — and won't — do with your data.
These aren't aspirations. Each is implemented in the deployment, written into the DPA, and falsifiable under audit. If any line below doesn't hold, we've failed the contract.
Your data stays in your perimeter.
Single-tenant VPC or on-prem. Read-through only. No replication to Nodes-side storage. No vendor-owned copy of PII, résumés, or interview audio. The deployment boundary is the contract boundary.
No training on your data without written consent.
Retrains are per-customer, in-VPC, on customer-consented data only. No cross-tenant pooling. No shared foundation updates from your corpus. Your records train your model, and only your model.
SOC 2 is what we hold. Everything else is operational alignment.
HIPAA, GDPR, NYDFS, ISO — your own program covers the deployment. We don't claim certifications we don't hold, and we don't list any as "in progress." The full matrix is in Security & Compliance.
Decision traces are reproducible seven years out.
Every score, every admin action, every retrain — hashed, signed, and streamed to your SIEM. A decision made today can be replayed in 2033 with the same inputs and the same outputs.
One production reference. Expanding.
We have one Fortune 500 deployment live today — 1,691 agents, 10 months in production, expanded two quarters ahead of schedule. The architecture is vertical-agnostic; the reference list isn't yet. We won't quote a corpus size for a deployment that doesn't exist.
30 minutes. Your data. No pitch deck.
We'll show you what's in your data before you commit to anything. Not a demo — a read on your hiring signal using your own production records. Your number is sitting in your data right now.