Skip to main content

Deployment Targets for AI Products

An AI Product must declare its deployment targets — the environments in which it can be run or accessed.
This ensures consumers know where and how the product can operate.

Why Deployment Targets Matter​

  • Transparency → Consumers understand where the product can run (cloud, edge, on-prem).
  • Portability → Enables consistent packaging across different environments.
  • Governance → Risk and compliance obligations differ across deployment types.
  • Efficiency → Deployment context influences resource requirements and cost models.

Deployment Target Categories​

1. Cloud​

  • Hosted in a public or private cloud environment.
  • Accessible via APIs or managed services.
  • Attributes: region, provider (AWS, GCP, Azure, etc.), SLAs, compliance certifications.

2. On-Premises / Private Data Center​

  • Deployed within an enterprise-controlled infrastructure.
  • Often chosen for data residency, security, or compliance reasons.
  • Attributes: hardware requirements, virtualization/containerization, integration needs.

3. Edge​

  • Runs on edge devices close to the data source (e.g., mobile devices, IoT sensors, cameras).
  • Attributes: device class (mobile, embedded, gateway), latency constraints, offline support.

4. Hybrid / Multi-Cloud​

  • Supports multiple environments simultaneously.
  • Attributes: interoperability requirements, federation models, failover strategies.

5. Agent Runtime​

  • Deployed inside an agent ecosystem or orchestration framework.
  • Attributes: level of autonomy, coordination model, communication protocol.
  • Particularly relevant for multi-agent systems.

Required Declarations​

Every AI Product must document:

  • Supported targets (cloud, on-prem, edge, hybrid, agent runtime).
  • Resource requirements (CPU, GPU, TPU, memory, storage).
  • Dependencies (libraries, runtime environments, orchestration frameworks).
  • Compliance constraints (data residency, security certifications, regional restrictions).

Example​

Vision Model API

  • Deployment Targets: Cloud (AWS, GCP), Edge (NVIDIA Jetson).
  • Resource Requirements: GPU w/ CUDA support.
  • Compliance Constraints: Must run in EU regions for GDPR compliance.

Summary​

  • Deployment targets describe where an AI Product can run.
  • Categories include Cloud, On-Prem, Edge, Hybrid, and Agent Runtime.
  • Products must declare resource requirements, dependencies, and compliance needs.

Principle: A true AI Product must declare its deployment targets so consumers can evaluate suitability and compliance before adoption.