Discovery call
30 minutes, no slides. We map the integration surface, the security constraints, and the people who need to sign off.
Self-Hosted AI Products
AI that runs on your terms
Enterprise-grade AI products you deploy on your own infrastructure. Full data sovereignty. No vendor lock-in. From the team that builds production AI systems.
Two products. Built to run on your infrastructure.
Lumen and Notetaker bring AI into your team’s daily work without sending a byte of your data to a vendor cloud. Self-hosted, audited, and yours to operate.
How we deploy on your infrastructure.
Self-hosted enterprise software lives or dies on the deployment story. We’ve made ours boring on purpose.
Security review
We answer your CISO’s questions in writing: architecture diagram, data-flow doc, threat model, and a compliance matrix tailored to your environment.
Pilot install
Containerized deployment to a staging environment in your VPC. Two weeks median, one engineer, yours stays on call, ours brings the runbook.
Integration
SSO, SCIM, your identity provider, your storage. Connectors to Slack, Teams, Google Workspace, M365, Confluence, Jira on day one.
Production rollout
We hand off a runbook, train your operators, and stay on call for the first 90 days. After that it’s yours to operate, no vendor dependency.
Why self-hosted, not SaaS.
The cloud-AI pattern doesn’t fit teams that own their data. We built for the ones who can’t compromise on residency, control, or audit.
01
Your Data Never Leaves
Everything runs in your network. No vendor cloud, no external API calls, no shadow data flows. Air-gap capable for the regulated cases.
02
Bring Your Own Models
Use ours, use OpenAI/Anthropic, or use your own self-hosted Llama, Mistral, or Particula models. The product is model-agnostic by design.
03
SSO, SCIM, RBAC
Plugs into your identity stack on day one. SAML/OIDC, SCIM provisioning, audit logs that go where your security team already looks.
04
Air-Gap Capable
For the regulated industries that need it: full offline operation, zero telemetry, no phone-home, no surprise outbound traffic.
05
No Lock-In
Source-available core. Data export at any time. Optional code escrow on enterprise contracts. If we go away, you keep running.
How we compare
Particula products vs SaaS vs in-house build.
Particula products
Self-hosted on your infrastructure
Source-available, no vendor lock-in
SSO, SCIM, RBAC out of the box
Two-week pilot install, named engineer
Flat licensing, no per-call surprises
SaaS competitors
Your data lives on vendor servers
Closed source, vendor roadmap
Auth bolted on after the fact
Sign up and hope it fits your workflow
Per-seat pricing that scales against you
Build in-house
12–18 months to ship v1
Compounding ML platform debt
Distracts from core product work
Hard to attract the right engineers
Maintenance burden in perpetuity
Why we built these
Every team I work with hits the same wall: the AI tools they want are SaaS-only, and the data they want to use is the data they can’t send to a vendor. So we built tools that run where your data already lives. That’s the product.
Sebastian MondragonFounder, Particula Tech
Frequently asked
On the infrastructure you choose. We deploy into your VPC, your on-prem cluster, or an air-gapped environment. We never see your tenant data.
Start a Conversation