Edge computing is the future where the volume of data is exploding and the need for secure, low-latency results proves to be more important in every domain. Think of any intelligent solution in any domain: Energy, medical, or retail, you could see that the need to get results quickly and efficiently is paramount. That’s one key reason why more solutions are deployed to the Edge; for fast, secure, and more efficient computing.
There are great opportunities in the IoT space generally, and specifically on the Edge. However, the challenges are also immense. Particularly around security, device Management, and deployments to the edge device, and here is where Azure Percept can be a game changer.
What is Azure Percept?
Azure Percept is a comprehensive, secure, and easy-to-use platform for creating AI solutions on the Edge. Azure Percept is a new type of devices, one that is designed for the future. Up until now, you can get an IoT device, and you need to manage and secure the device yourself, then you need to integrate the device with some AI services to be able to deploy machine learning models so that you can get the device to do something useful, and this is where Azure Percept can be a game-changer.
Azure Percept is designed to natively and seamlessly integrate with Azure IoT and AI Services to give you a head start by doing most of the common plumbing work that you might need to do, and enables you to focus on your business domain and business challenges.
Azure Percept is a new type of device that brings Cloud and Edge together and provides a seamless experience to manage, train, and deploy models quickly, securely, and efficiently. Azure Percept provides a platform to get started quickly, pilot, then productionise and scale.
Azure Percept has a few parts:
- Azure Percept Developer Kit (DK): A flexible, and intelligently designed development kit that includes solid hardware with vision and speech capabilities. The device can be used for quick prototyping and allows for piloting solutions securely in minutes.
- Services and workflows that enable you to quickly manage, secure, deploy and scale AI models and Edge modules.
- AI hardware reference design and certification programs. This effectively codifies security and best practices into your hardware designs.
- Azure Percept Studio which is an online portal to allow you to easily manage your device, and deploy models and services over-the-air.
Why Azure Percept?
There are a few reasons why you should consider building your AI solution on Azure Percept.
- Solid, and flexible hardware design that comes with vision and speech capabilities, and built-in hardware-acceleration for better performance on the Edge.
- Secure by design with a hardware security module to ensure end-to-end security.
- Native integration with Azure IoT, and AI services for a simpler workflow of managing, training, and deploying models.
- Azure Percept integrates natively with Azure Percept Studio and Azure Percept Audio to enable you to manage and deploy to the device from the browser (within the Azure Portal).
- Integration with common Open Source AI tools and frameworks like TensorFlow and ONNX.
How about Security?
Security continues to be a major challenge for IoT and AI solution. The number of cyber-security breaches is on the rise, with greater sophistication in tools, tactics, and access. That’s why its critical to get security baked into AI and Edge solutions from the start.
Azure Percept uses a zero-trust security model to safe-guard your solution end-to-end, from silicon to cloud. It comes with a Trusted Platform Module (TPM 2.0). TPM is an industry-wide ISO standard from the Trusted Computing Group, which can be used along with Azure services like Azure Device Provisioning Service (DPS) for secure onboarding and management of the device. This not only ensure the Edge device, but also protect the privacy-sensitive data on it. Using secure hardware-backed encryption enables data security in transit and at-rest.
The Deployment Model
If you have built any AI solution, then you will be familiar with the challenges of versioning, labeling, and reproducing models. This is even more challenging when your AI solution is based at the Edge as you do not always have physical access to the device. These are some of the challenges that are made a lot easier with the use of Azure Percept Studio. Azure Percept enables:
- Simple onboarding of device and modules. I was able to follow the prompts and onboard my device within minutes.
- Integrated experience where the boundaries between Azure services, as well as between cloud and Edge are fading.
- Easy-to-use no-code flow and interfaces for quick training and deployment of models, while at the same time providing the option to create advanced flows if needed. In my case, I created and deployed my model in under one hour.
- Simplified models and modules management on the Edge for accelerated prototyping.
I truly think that Azure Percept can be a game-changer, and it is signaling the rise of a new type of devices where the boundaries between cloud and edge are fading. As we have seen, Azure Percept provides a comprehensive, secure, and easy-to-use platform for piloting new AI solutions on the edge with state-of-the-art hardware, security, and deployment tools.