AI edge engineer

AI Engineer
Data Scientist
Azure Notebooks
Cloud Shell
Container Instances
Container Registry
IoT Edge
IoT Hub
Machine Learning Service
Azure Resource Manager
Virtual Machines

The interplay between AI, cloud, and edge is a rapidly evolving domain. Currently, many IoT solutions are based on basic telemetry. The telemetry function captures data from edge devices and stores it in a data store. Our approach extends beyond basic telemetry. We aim to model problems in the real world through machine learning and deep learning algorithms and implement the model through AI and Cloud on to edge devices. The model is trained in the cloud and deployed on the edge device. The deployment to the edge provides a feedback loop to improve the business process (digital transformation).

In this learning path, we take an interdisciplinary engineering approach. We aspire to create a standard template for many complex areas for deployment of AI on edge devices such as Drones, Autonomous vehicles etc. The learning path presents implementation strategies for an evolving landscape of complex AI applications. Containers are central to this approach. When deployed to edge devices, containers can encapsulate deployment environments for a range of diverse hardware. CICD (Continuous integration - continuous deployment) is a logical extension to deploying containers on edge devices. In future modules in this learning path, we may include other techniques such as serverless computing and deployment on Microcontroller Units.

The engineering-led approach underpins themes / pedagogies for engineering education such as 

  • Systems thinking
  • Experimentation and Problem solving
  • Improving through experimentation
  • Deployment and analysis through testing
  • Impact on other engineering domains
  • Forecasting behaviour of a component or system
  • Design considerations
  • Working within constraints/tolerances and specific operating conditions – for example, device constraints
  • Safety and security considerations
  • Building tools which help to create the solution
  • Improving processes - Using edge(IoT) to provide an analytics feedback loop to the business process to drive processes
  • The societal impact of engineering
  • The aesthetical impact of design and engineering
  • Deployments at scale
  • Solving complex business problems by an end-to-end deployment of AI, edge, and cloud.

Ultimately, AI, cloud, and edge technologies deployed as containers in CICD mode can transform whole industries by creating an industry-specific, self-learning ecosystem spanning the entire value chain. We aspire to design such a set of templates/methodologies for the deployment of AI to edge devices in the context of the cloud. In this learning path, you will:

  • Learn about creating solutions using IoT and the cloud
  • Understand the process of deploying IoT based solutions on edge devices
  • Learn the process of implementing models to edge devices using containers
  • Explore the use of DevOps for edge devices

Produced in partnership with the University of Oxford – Ajit Jaokar Artificial Intelligence:Cloud and Edge Implementations course.



Modules in this learning path

Explain the significance of Azure IoT and the problems it solves. Describe Azure IoT components and explain how you combine them to solve IoT solutions, which create value for enterprises.

Assess the characteristics of Azure IoT Hub and determine scenarios when to use IoT Hub.

Explain the essential characteristics of the IoT Edge and the functionality of the IoT Edge components (modules, runtime, and cloud interface). Characterize the types of problems that you can solve with IoT Edge. Describe how the elements of IoT Edge can be combined to solve the problem of deploying IoT applications in the cloud.

Deploy a pre-built temperature simulator module to the edge using a container. The pre-built module will be deployed to an IoT edge device. You will check that the module was successfully created and deployed to the edge. You will view the simulated data from the deployed module.

Deploy a trained machine learning module to the edge using a container. The machine learning module you create will be deployed to an IoT Edge device. You'll check that your container image was successfully created and stored in the Azure container registry. You'll view the data from the deployed module from the IoT Edge.

Assess the characteristics of Azure Functions for IoT. Describe the function of triggers and bindings and show how you combine them to create a scalable IoT solution. Describe the benefits of using cloud infrastructure to rapidly deploy IoT applications with Azure Functions.

Create and deploy an Azure function to make a language translation IoT device. The function will use Cognitive Speech Service. Your device will record a voice in a foreign language and convert the speech to a target language.

Implement a cognitive service for performing language detection on an IoT Edge device. Describe the components and steps for implementing a cognitive service on an IoT device

Analyze the significance of MLOps in the development and deployment of machine learning models for IoT Edge. Describe the components of the MLOps pipeline and show how you can combine them to create models that can be retrained automatically for IoT Edge devices.

Define a solution for smoke testing for virtual IoT Edge devices. Your solution will employ a CI/CD (Continuous Integration/ Continuous Deployment) strategy using Azure DevOps, Azure Pipelines, and Azure Application Insights on a Kubernetes cluster.

Determine the types of business problems that can be solved using Azure Sphere. Explain the capabilities and the components (microcontroller unit, operating system, cloud-based security service) for the Azure Sphere. Describe how the components provide a secure platform to develop, deploy, and maintain secure internet connected IoT solutions.

Implement a neural network model for performing real-time image classification on a secured, internet-connected microcontroller-based device (Azure Sphere). Describe the components and steps for implementing a pre-trained image classification model on Azure Sphere.