Platform · Deployment · US

Compare AI Deployment Options and Operating Responsibilities

The same container images run in every deployment model - what changes is who manages infrastructure, access, updates, monitoring, and recovery. This page walks through those responsibility boundaries for a US evaluation.

Overview

LuMay ships as Docker containers, and the same images run across every deployment model. What changes between models is not the application code - it's who manages the surrounding infrastructure, who has operational access, and where data lives. This page does not make any claim about a specific US hosting location, data residency guarantee, or air-gap capability beyond what the deployment model itself structurally supports - confirm specific requirements directly before committing to a model.

Deployment choice does not change the core

LuMay SaaS

Management API
Voice Agent
Data Services
Observability
Managed

Customer Cloud

Management API
Voice Agent
Data Services
Observability
BYOC

Private Cloud

Management API
Voice Agent
Data Services
Observability
VPC

On-Prem K8s

Management API
Voice Agent
Data Services
Observability
Data Center

The repeated shape communicates portability more strongly than any claim.

Deployment Models

ModelWho manages infraWhere data livesBest for
LuMay SaaSLuMayLuMay's cloud tenancyFast start; teams that don't want to manage infrastructure
Customer cloud (BYOC)LuMay manages apps; you own the subscriptionYour cloud subscriptionData-residency requirements; shared responsibility model
Private cloud / VPCYour ops team or LuMayYour VPC with network isolationRegulated industries with strict network controls
On-prem KubernetesYour ops teamYour data centerFull air-gap capability; defense, government, high-security environments
HybridSplit between LuMay and customerManagement plane in cloud; data plane on-premEnterprises with mixed environments

Responsibility Boundaries

  • Infrastructure provisioning - LuMay in SaaS; split in BYOC/private cloud; your team in on-prem.
  • Operational access - who can access production systems, and is that access logged and auditable, changes by model.
  • Updates & patching - confirm the update cadence and who approves a production rollout in your chosen model.
  • Monitoring - the same observability stack (OpenTelemetry, Prometheus, Loki) runs in every model; who watches the dashboards is a staffing decision on your side for self-hosted models.
  • Recovery - ask directly about backup cadence and recovery time for your specific model before committing - this page does not assert a specific recovery-time figure.

Illustrative Example

The following is illustrative, not a specific customer commitment.

A team comparing managed SaaS hosting against a customer-controlled private-cloud deployment lists out, for each candidate model: who provisions the database, who has production shell access, how a security patch gets applied, and who is paged if the voice service goes down at 2am. Writing this comparison out explicitly, rather than assuming it's the same across models, is what surfaces the real operational trade-off between the two options.

FAQ

Does moving from SaaS to on-prem require re-architecting the application?

No - the same container images run in every model; what changes is infrastructure ownership, not the application.

Can we get a specific data-residency or air-gap guarantee for a US deployment?

Confirm specific residency, air-gap, or compliance requirements directly for your deployment model - this page does not assert a specific guarantee.

Who manages scaling in each model?

In SaaS and BYOC, Azure Container Apps auto-scaling applies; in private cloud/on-prem, your Kubernetes configuration determines scaling behavior.

Is pricing different by deployment model?

Pricing is outside the scope of this page - see /us/pricing for pricing information.

What observability do we get in a self-hosted model?

The same OpenTelemetry/Prometheus/Loki stack runs in every model - see Analytics & ROI.

Which deployment model fits your environment?

Book a 30-minute deployment review. We'll assess your network and operational constraints and walk through the responsibility split for each model.