Now in Public: Mobilint ARIES Demo Ecosystem and Boilerplate Repositories

We have an exciting announcement: our suite of demo programs for ARIES-based hardware is now available on GitHub!

Access the Repositories: https://github.com/orgs/mobilint/repositories

These demos, spanning applications using vision models to multi-instance large language models, provide an accessible entry point and a versatile "playground" for developers beginning their journey with Mobilint hardware.

We recognize that porting neural network models to high-performance edge deployments can involve significant friction, and making projects zero-to-one on a new environment could be a challenge. At Mobilint, we believe that providing hardware is only half of the equation; providing the architectural blueprint for deployment is the other.

Here is a closer look at what you can expect from these new repositories.

An Intuitive and Simplified Boilerplate for NPU Integration

These repositories are engineered to serve as functional boilerplates for our users and partners. By providing standardized code for qb Runtime utilization, we aim to offer an intuitive building experience and significantly reduce the time-to-prototype for engineers developing complex systems.

Key technical advantages of the new demo ecosystem include:

• Unified Python Codebase: For Python-based implementations, we have structured the code so the exact same script can execute across both GPU and Mobilint NPU environments. This allows for seamless verification on existing GPU workstations before shifting the workload to ARIES for production-grade efficiency.

• Cross-Platform Parity: Recognizing the diverse environments in industrial and commercial AI, all demos are verified for native operation on both Linux and Windows.

• Optimized Resource Management: The repositories provide clear examples of how to manage NPU device numbers, core modes, and target cores, ensuring engineers can maximize hardware utilization from day one.

  
  

Demos to Explore: From Object Detection to Multi-LLM Serving

By moving our demo assets to a public GitHub environment, we have streamlined the installation process for external networks. The Mobilint hardware ecosystem is now just a git clone away.

Here are some key repositories at a glance (with more coming soon):

Repository Name	Primary Tech Stack	Application Focus
aries-vision-transformer-demo	TypeScript, Next.js, Python	High-efficiency Vision Transformer deployment
aries-llm-demo	TypeScript, Next.js, Python	Chatbot-type application with a selected list of Large Language Models
aries-multi-llm-demo	TypeScript, Next.js, Python	Multi-instance Large Language Models supporting up to 16 concurrent conversations
aries-vlm-demo	TypeScript, Next.js, Python	Vision-Language (VLM) understanding on the edge
aries-ai-assistant-demo	TypeScript, Next.js, Python	Integrated AI assistant and conversational logic
aries-multi-channel-demo	C++	Concurrent multi-stream video processing
aries-fire-detection-demo	C++	Real-time fire and smoke detection for safety
aries-weapon-detection-demo	C++	Specialized object detection for security and surveillance

For example, here's the view of what you would get if you had an MLA400 PCIe Card and our Multi-LLM demo.

Engineering Excellence through Accessibility

Mobilint is committed to lowering the barrier to entry for high-performance Edge AI. By providing open, well-documented source code, we empower the engineering community to move beyond theoretical performance and into scalable, real-world applications.

If you or your team have recently started working with Mobilint’s ARIES-powered hardware, we invite you to clone the repositories, explore the qb Runtime implementation, and use these assets as the foundation for your next AI solution.

---

Beomsu Lim

Senior Research Engineer, AI Solutions Team