This post has been republished via RSS; it originally appeared at: New blog articles in Microsoft Tech Community.
Guest post by Ajit Jaokar is the course director for the Artificial intelligence: Cloud and Edge implementations course at the University of Oxford
Background
In this post, I would like to share my experiences in teaching AI Edge Engineering concepts at the Artificial intelligence: Cloud and Edge implementations course at the University of Oxford. I am the course director and the lead tutor for this course at the #universityofoxford
I recognise that this is a complex and a pioneering area. We welcome comments and feedback (especially from other Universities and Research Institutions globally) who are implementing similar strategies. Our code and modules are shared in Open Source (links in this article below).
Motivations
In the recently concluded world economic forum in Davos, the impact of AI on jobs was a big theme.
The WEF concluded that tech skills will dominate for jobs of the future.
The dominance of tech skills for the future is not a surprise.
However, there is a caveat.
All over the world, developers will need to upskill to the world of AI.
In many ways, this is an even more urgent need because first AI systems have to be built before these systems have an impact on the wider employment.
And we need developers to upskill and learn AI to build these new systems.
The stakes are high. Nations who can bridge this gap will leapfrog their economies in a big way. The problem is recognised by policy makers. In the UK, the government has introduced a national retraining scheme for the tech skills gap in the face of AI growth.
The AI Edge Engineer concept
Our course is based on the concept of the “AI Edge Engineer”
As the title suggest, the “Artificial intelligence: Cloud and Edge implementations” course spans the Cloud and the Edge for implementing AI and ML algorithms.
The course is engineering led.
From a technical standpoint, this term ‘engineering led’ has a specific meaning.
For us, that means, we consider the full pipeline for model building and model deployment i.e. MLOps (AI, DevOps) for Edge devices
We cover the full stack from the cloud to the edge using containers.
From a skills standpoint, this scope actually encompasses three jobs relating to AI: the data engineer, the data scientist and the devops engineer.
The pedagogical challenges and benefits
The scope presents a challenge from a pedagogical perspective.
The scope and complexity of the subject is vast, and AI is rapidly evolving
Currently, we have deployed the following modules
- IoT Hub
- IoT Edge
- Deploy a prebuilt module to Edge devices
- Build and deploy a module to the Edge
We are also planning to deploy modules covering
- Serverless
- Sphere
- CI/CD
for edge devices
In future, we are planning for the edge
- Kubernetes
- Computer Vision
- Digital Twin
We use Python (tensorflow and keras) for all our coding and we are deploying the code and modules in Microsoft Learn. This Microsoft Learn format enables task-based learning for topics which can be combined into logically interconnected learning paths. This helps overcome some of the teaching challenges. This approach helps us to avoid cognitive dependencies and cognitive overload for such a complex topic. Our code and modules are deployed at the AI Edge Engineer learning path
Benefits for acquiring new skills in AI
While the scope of the modules is complex, the emphasis on MLOps (AI and DevOps) for Edge devices has multiple benefits for participants. MLOps helps them to leverage their existing skills and also then co-relate to new skills. For example, if they have a background in ETL (Extract, Transform, Load) – they could start as a data engineer and then learn about how that data gets used in data science and devops. Participants also get Cloud and Edge programming skills. This full stack experience helps them to transition to jobs in larger enterprises with similar architectures.
Conclusion
It is still early days, but the response has bee positive. Course participants are building and deploying modules on edge devices. We shall continue to share our insights and code as LEARN modules. We welcome insights and feedback from other educators globally. You can contact me via LinkedIn Ajit Jaokar.