# Metro AI Suite Developer Workflow

**Metro AI Suite** is a developer stack for edge vision AI, generative AI, and
physical AI.

This section describes the Metro AI Suite workflow for developers. See
the repository-wide suite overview and application index in
[Metro AI Suite](https://docs.openedgeplatform.intel.com/dev/ai-suite-metro.html);
application-specific details remain in each application's documentation.

## Create and Acquire Models

You have the following options:

- Build a model using visual prompting with [Intel® Geti™](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/geti.html)
  software.
- Select a model from the [Metro Analytics Catalog](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/ai-suite-metro/metro-analytics-catalog.htmlg)
  and [OpenVINO Model Collections](https://huggingface.co/OpenVINO/models)
  hosted in the Hugging Face hub.
- Bring your own model. See [Model Convertion Guides](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/dlstreamer/dev_guide/model_preparation.html) and [Agent Skills](https://github.com/open-edge-platform/skills).

## Choose Platform

Intel recommends the following steps:

1. Evaluate Hardware: Evaluate your hardware using the
   [Visual Pipeline and Platform Evaluation Tool](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/visual-pipeline-and-platform-evaluation-tool/index.html).

2. Search AI-Ready Qualified Systems: Search for AI-ready qualified systems in the
   [Recommended Hardware Catalog](https://builders.intel.com/ecosystem-engagement/solution-hub/edge-ai-catalog/partner-spotlight).

3. Access Pre-Benchmarked Results from [Metro AI Suite Benchmarks](https://cdrdv2.intel.com/v1/dl/getContent/856310?explicitVersion=true) and
   [OpenVINO benchmarks](https://docs.openvino.ai/2026/about-openvino/performance-benchmarks.html).

## Build Faster

### Jumpstart Solutions with AI Sample Applications and Blueprints

For example, use [Live Video Caption & Search](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/live-video-captioning/index.html) and [Video Search and Summarization](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/video-search-and-summarization/index.html).

You can explore other sample applications and blueprints in the
[Metro AI Suite](https://docs.openedgeplatform.intel.com/dev/ai-suite-metro.html).

### Build Production-Grade AI Pipelines Modularly

Use [Vision and AI SDKs](https://docs.openedgeplatform.intel.com/dev/OEP-articles/oep-sdk-manager/index.html), which include:
- Microservices
- Libraries
- Tools
- Third-party components for video and AI workloads

### Build Software with Coding Agents

Use [Metro AI App Builder](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/metro-prompt-library/index.html) and [Open Edge Platform Skills](https://github.com/open-edge-platform/skills).

## Add Capabilities

### Make Solutions Autonomous with Agentic AI

- Use a [Live Video Alert Agent](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/live-video-alert-agent/index.html).
- Use a [Smart Route Planning Agent](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/smart-route-planning-agent/index.html).
- Use [Agentic Predictive Maintenance](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/ai-suite-metro/agentic-predictive-maintenance-pipeline/index.html).

### Integrate AI with a Video Management System (VMS)

- See [VMS Integrations](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/vms-adapter-plugin/index.html) and [Smart NVR](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/smart-nvr/index.html).

### Bring AI into the Physical World

- Use [Scenescape](https://docs.openedgeplatform.intel.com/dev/scenescape/index.html).
- Use [Smart Building Digital Twin](https://docs.openedgeplatform.intel.com/dev/edge-ai-suites/ai-suite-metro/app-blueprint-building-twin.html).

## Optimize

Maximize performance for edge deployment:

- Use the [DL Streamer Pipeline Optimization Tool](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/dlstreamer/dev_guide/optimizer.html).
- Use the [Visual Pipeline and Platform Evaluation Tool](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/visual-pipeline-and-platform-evaluation-tool/index.html).
- Refer to [Performance Blueprints](https://builders.intel.com/ecosystem-engagement/solution-hub/systems/edge-systems-qualification/ai-edge-systems#vertical-specific-workloads).
- Use the [Deep Learning Streamer (DL Streamer)](https://docs.openedgeplatform.intel.com/dev/edge-ai-libraries/dlstreamer/index.html).

## Migration and Modernization

The suite supports migration from NVIDIA-based stacks to Intel-native execution.
See the [guide](https://docs.openedgeplatform.intel.com/dev/OEP-articles/nvidia-migration.html)
for more information.