# Deploy Tinkerblox UltraEdge HPC-I for AI and mixed workloads on Arm

## In this learning path

- [Introduction](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/)
- [Understand UltraEdge HPC-I architecture for edge AI and mixed workloads](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/background/)
- [Provision a Google Axion C4A VM for Yocto image builds on Arm](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/instance/)
- [Build and install Yocto images for NXP S32G-VNP-GLDBOX3 with UltraEdge](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/yocto_installation/)
- [Install UltraEdge on Debian and Ubuntu for Edge AI workloads](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/debian_installation/)
- [Run and manage UltraEdge HPC-I for AI and mixed workloads on Arm](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/running/)
- [Next Steps](https://learn.arm.com/learning-paths/cross-platform/tinkerblox_ultraedge/_next-steps/)

## About this Learning Path

| Skill level: | Advanced |
|--------------|----------|
| Reading time: | 1 hr |
| Last updated: | 05 Aug 2026 |

| Author: | Tinkerblox |
|---------|-------------|
| Arm IP: | [Neoverse](https://support.arm.com/?tab=compute-ip&Product%20Type=Infrastructure%20Processors) |
| Tags: | [Containers and Virtualization](https://learn.arm.com/tag/containers-and-virtualization), [Google Cloud](https://learn.arm.com/tag/google-cloud), [Linux](https://learn.arm.com/tag/linux), [other](https://learn.arm.com/tag/other), [Tinkerblox](https://learn.arm.com/tag/tinkerblox) |

### Who is this for?
This is an advanced topic for business, R&D, and engineering teams seeking to optimize CPU and GPU infrastructure utilization while reducing total cost of ownership on edge and constrained environments. It's ideal for innovation and development teams building next-generation AI workloads using alternative runtime environments and packaging technologies.

### What will you learn?
Upon completion of this Learning Path, you will be able to:
- Understand the layered architecture of UltraEdge core, boost, and prime
- Build applications using the UltraEdge MicroStack
- Deploy the MicroPacs on Linux-based compute systems and scale to cloud or data-center environments
- Optimize performance for edge-cloud scenarios, enabling near real-time data flows

### Prerequisites
Before starting, you will need the following:
- Experience using Linux on embedded or SBC platforms
- Understanding of container runtimes (containerd) and CNI networking
- Basic knowledge of communication protocols (MQTT, HTTP, and others)
- Familiarity with edge-cloud architectures and data-flow orchestration

### Summary
You’ll explore Tinkerblox UltraEdge, an edge-native execution fabric for AI and mixed workloads on Arm platforms. First, you’ll review the UltraEdge architecture, provision a Google Axion C4A virtual machine for an `arm64` Yocto build, and install the UltraEdge agent and MicroPac tooling on Debian or Ubuntu. Then, you’ll build and deploy MicroPac workloads, create a Yocto image for an `NXP S32G-VNP-GLDBOX3` board, and manage services with the Tinkerblox and MicroBoost CLIs. The workflows cover both package-based Linux installations and Yocto-based device deployments.

### Frequently asked questions

#### What is the purpose of UltraEdge in this Learning Path?
UltraEdge provides an edge-native execution fabric for AI and mixed workloads. Review its layered architecture and see how MicroStack and MicroPac support workload packaging and deployment on Arm-based systems.

#### Why is a Google Axion C4A virtual machine used for the Yocto build?
The C4A virtual machine provides the environment for building the image for the target NXP board.

#### Can UltraEdge be installed on Debian or Ubuntu?
Yes. Install and activate the UltraEdge agent on Debian or Ubuntu, then use MicroPac tooling to define, build, validate, and install workloads.

#### What are the requirements for building the Yocto image for the NXP board?
Use the `NXP S32G-VNP-GLDBOX3` platform with BSP `38.0`, an AArch64 Ubuntu build host, the required Yocto layers, and the `meta-edgeblox.zip` layer requested from Tinkerblox support. Yocto builds can take several hours depending on the available resources.

#### How do I manage workloads after installing UltraEdge?
Use the Tinkerblox CLI on Debian or Ubuntu, or the MicroBoost CLI on the Yocto-based device, to install, start, stop, monitor, and diagnose MicroPac-based workloads. You can also use the CLIs to inspect system state and troubleshoot common connection and architecture issues.