Model selection
Select the right model for the task and its sensitivity.

Metis ARX provides the private AI foundation. AERIX extends it through policy-controlled access to cloud capabilities, with data governance at every layer.
Book a demo
A central gateway for approved private and cloud AI models. AERIX applies routing policies, access controls and usage visibility across connected applications.
Select the right model for the task and its sensitivity.
Switch models within approved policies as needs change.
Keep client data local; route public-safe work to Cloud AI.
Use approved fallback routes when a model is unavailable.
Use cloud LLMs and private models without a single-model dependency.
Mask sensitive information and apply access policies before routing.
Continuity controls reduce disruption; availability depends on deployment.
Observe model usage, decisions, risk events and cost allocation.
PRIVATE AI · THE CORE
Your data. Your infrastructure. Your control.
ONE SECURE GATEWAY
CLOUD AI · THE EXTENSION
Public innovation. On your terms.
| Dimension | Cloud AI | Private AI · ARX |
|---|---|---|
| Strength | Scale, speed, latest capabilities | Privacy, control, custom knowledge |
| Workload | General knowledge, language fluency, public research | Raw customer data, contracts, internal documents |
| Models & data | Cloud LLMs and specialist models | Private model, private LLM, private data |
| Delivery | Public hyperscale cloud | On-premises appliance or private cloud |
Cloud AI: hyperscale intelligence, elastic compute, global knowledge
Secure hybrid-cloud bridge
Private compliance: air-gapped data control, regulatory rules, PII sovereignty, local audit
Enterprise risk handoff · human approval
| Layer | Cloud capability | Bridge control | On-premises control · ARX |
|---|---|---|---|
| Risk Governance | Vendor SLA · Risk registry | Risk handoff · Human approval | Deterministic rules · Local accountability |
| Governance & Visibility | Usage analytics · Service metrics | Audit log · DLP masking · FinOps meter | RBAC/ABAC · Local audit · Zero trust |
| Agent & Application | Public-safe language agents | Human-in-the-loop trigger | KYC / AML / Wealth agents |
| Orchestration & Tools | Elastic tool execution | Semantic router · Tool gateway | Sandbox runtime · Approved tools |
| Data & Memory | Public knowledge retrieval | RAG routing · PII masking | Vector database · Private memory |
| Foundation Model | Cloud LLMs · Specialist models | Model evaluation · Bias testing · Fallback | Local private model · Private fine-tuning |
| Infrastructure | Enterprise cloud · Elastic compute | Encrypted tunnel · Policy boundary | On-premises · Secure enclave |
Your private AI foundation for sensitive data and regulated workflows, deployed on-premises or in an air-gapped environment.
Offline inference on enterprise hardware, with no external API calls during inference.
Every query, output and document access is logged with user IDs and encrypted timestamps.
Granular role-based access by department or branch.
Answers derive from verified internal sources, with page and paragraph citations. If no source exists, the system says so and does not guess.
Vectors and model parameters remain on-premises.
Plan capacity across hardware with MET (Metis Equivalent Token), and align module costs with business needs through BCU (Business Capability Units). Discuss a configuration tailored to your workload.
LS-Matrix provides hybrid LLM core-space alignment between private and cloud capabilities. It supports consistent orchestration while keeping sensitive knowledge within the private core.
Governance is part of the operating model, not an afterthought.
Trace each request from model selection to output and approval.
Understand usage and allocation by workflow and module.
Apply hard rules and human gates before decisions take effect.
Private control remains the foundation of the platform.
ARX is the private AI core. AERIX extends it with policy-governed access to approved cloud AI. Sensitive workflows retain source citations, human approval and audit visibility.
Discuss your operational priorities, deployment requirements and evaluation criteria with our team.

