Policy before routing
Every request is classified for sensitivity and task type. Residency rules are enforced before any model is considered — and violations fail closed, never silently downgrade.
Cardix
The cardinal axis for your AI
Cardix sits inside your network as an intelligent routing layer — so every request reaches the right model under your rules, not ours.
Intelligent routing for hybrid AI — inside your perimeter.
Why Cardix
Three pillars of in-perimeter AI routing
Right model, every request
Cardix selects across your full catalog — on-prem, private cloud, and approved APIs — optimizing for quality and cost under your constraints.
Gets smarter over time
Cardix learns from your traffic which models excel at which tasks. Intelligence stays on your infrastructure; your data never reaches us.
Who we serve
Built for the teams running hybrid AI
Security, platform, FinOps, and regulated verticals — each with a distinct pain, each solved by policy-first routing inside your perimeter.
Security & compliance
Data leaving the perimeter before routing is decided
Policy before routing; fail closed on every violation.
Platform & AI infra
Mixed model fleet with no single routing endpoint
One layer across on-prem and approved cloud APIs.
FinOps & leadership
Defaulting every query to a frontier model burns budget
Right model every request; measurable savings over time.
Regulated verticals
Hybrid AI is mandatory but sovereignty cannot bend
Built for finance, healthcare, legal, and platform teams.
Industries
Where Cardix fits
Regulated enterprises running AI across on-prem and cloud — without compromising data sovereignty.
Financial services
Client portfolios and trading data must stay on-prem; summarization can use cloud when policy allows.
Cardix enforces matter-level routing without re-architecting existing AI integrations.
Healthcare
Clinical notes contain PHI that cannot leave the hospital network.
Restricted-classified requests route only to on-prem models; violations are blocked and logged.
Legal & professional services
Matter sensitivity varies; partners need confidence data is handled correctly.
Policy rules map data labels to allowed locations before any model sees the prompt.
Platform engineering
Team manages on-prem inference plus cloud APIs for different workloads.
One OpenAI-compatible endpoint; Cardix handles eligibility and selection across the fleet.
Ecosystem
Fits your stack
Cardix sits between your applications and your model backends — a single routing layer, no rip-and-replace.
OpenAI-compatible apps
Point your existing SDK at Cardix — no application rewrite.
On-prem inference
Route to models running inside your network boundary.
Private cloud models
Unified routing across your private deployments.
Approved cloud APIs
Dispatch only to backends you explicitly allow.
Your VPC / air-gap
Deploys where your data already lives.
Built for regulated enterprises
Hybrid AI that respects your perimeter — policy-first, fail-closed, and deployed on your infrastructure.
Policy before routingFail-closed enforcementYour data never reaches usDeploys inside your networkOpenAI-compatible
Clarity
What Cardix is not
Cardix is a focused routing intelligence layer — not a platform you have to rebuild around.
- Not a hosted SaaS that sees your prompts
- Not an inference platform or MLOps suite
- Not another chat UI
Get started
The cardinal axis for your AI
See how Cardix routes every AI request inside your perimeter — under your rules, not ours.