DEEPX has leveraged Network Optix’s AI Manager and the Open AI Accelerator eXchange (OAAX) standard to integrate its DX-M1 and DX-H1 NPUs into the Nx EVOS platform and broader Nx ecosystem.

Known for their exceptional power efficiency, compact architecture, and stable high-load performance, DEEPX NPUs deliver significant energy and cost advantages for edge-AI deployments.

Built on the OAAX framework, Nx AI Manager provides a frictionless path for AI accelerators to integrate and deploy at scale. The DEEPX integration demonstrates how hardware partners can achieve rapid onboarding, optimised performance, and scalable deployment without hardware-specific configurations or redevelopment.

“The Nx EVOS platform provided an ideal environment for integrating DEEPX’s high-performance, low-power NPUs,” said the DEEPX team. “The combination of Nx’s open architecture and DEEPX’s technology sets a new standard for edge AI video analytics.”

Integration Journey

Leveraging Nx AI Manager’s OAAX-based framework, DEEPX engineers integrated their proprietary SDK and runtime environment into the Nx EVOS platform through a single, standardised interface. The process was fast, transparent, and fully aligned with DEEPX’s modular SDK architecture.

Nx AI Manager’s comprehensive documentation and cloud-based workflow enabled quick setup and testing. Models can now be converted to DEEPX’s dxnn format and deployed directly via Nx Cloud across multiple servers or edge devices, without the need for manual configuration or system-specific tuning. 

“Nx AI Manager’s OAAX framework allowed us to port our SDK and runtime tools efficiently”, noted the DEEPX engineering team.  “The combination of Nx EVOS’s cloud system and OAAX made model conversion and deployment extremely fast and intuitive.”

This process demonstrates how Nx AI Manager enables a frictionless, hardware-agnostic AI pipeline, reducing engineering complexity while expanding compatibility across deployment environments.

Combined with DEEPX’s optimized SDK and power-efficient NPU architecture, the integration delivers real-time inference with more than ten times the power efficiency and reduced CPU utilization compared with conventional CPU- or GPU-based systems.

Optimisation and Collaboration

During development, DEEPX focused on maximising inference performance for multi-channel AI workloads. Using its dxtop hardware profiling tool alongside OAAX benchmarking utilities, DEEPX identified and resolved pipeline bottlenecks, ensuring stable, high-performance inference on Nx EVOS.

The team also implemented a CI/CD workflow to automate build, testing, and deployment, further accelerating integration and enhancing runtime reliability.

Close collaboration between DEEPX and the Network Optix engineering team played a key role in refining the runtime and achieving consistent performance under demanding, real-world conditions. 

“This project became a strong demonstration of synergy between DEEPX’s NPU optimisation expertise and Nx’s platform compatibility.”,  said the DEEPX engineering team.

The result is a repeatable, efficient integration process that other hardware vendors can follow when onboarding their accelerators through Nx AI Manager and OAAX.

Broader Value for Hardware Manufacturers

The DEEPX integration highlights the advantages of joining the Nx EVOS ecosystem for hardware and AI accelerator manufacturers. Through Nx AI Manager and OAAX, partners can bring their inference runtimes and SDKs into the Nx environment with minimal development effort, enabling:

• Faster time-to-integration through unified APIs and comprehensive documentation
• Hardware-agnostic AI model deployment without device-specific reconfiguration
• Scalable performance across servers, devices, and environments
• Access to a global developer ecosystem built around open standards

“Joining Nx’s ecosystem is more than an integration—it’s a strategic partnership”, said the DEEPX team. “Open standards like OAAX are the foundation of future scalability.”

For manufacturers, this means less engineering overhead, faster market access, and the ability to demonstrate unique hardware capabilities within a trusted global platform for intelligent video and edge AI applications.

To read more Network Optix news, click here.