How DevOps and AI can work together
Sometimes two separate things can combine to make something that's even better.
DevOps itself is an example of this, which is a combination of what used to be two disparate teams: development and operations. The combination of these two has resulted in a process that is more automated, repeatable and efficient while being able to deliver benefits to the business in a near continuous manner.
So what would happen if you combine DevOps with AI? When you think of DevOps, you don't naturally think of AI, and vice versa. However, when these two items come together, you can get something much better than either item on its own. There are two different ways that DevOps and AI can be joined. The first is where you apply AI to the DevOps process and the second is where you apply DevOps concepts to AI development.
Applying AI to the DevOps process
If DevOps is the union of people, process and technology to provide continuous value to the business, then what would happen if you overlaid AI on the DevOps process to automate some of the people and process parts. This is known as AIOps, which is a term that was first coined by Gartner in 2016.
The idea of AIOps is that if you could collect all the performance metrics, service tickets and usage patterns, you could use AI to identify areas of improvement and to perform root cause analysis. This would help IT develop a better functioning and performing application for the business and it would remove much of the guesswork out of the DevOps process, thus improving consistency.
Applying DevOps to AI development
The key purpose of DevOps is to have a development cycle that is more collaborative and cohesive, while providing incremental change in an automated and repeatable manner. These same concepts can be applied to AI models. What if, instead of having multiple data scientists working in isolation on local machines, you could have them work in a more iterative and collaborative way?
DevOps for AI promotes collaboration during AI model development, testing and refinement. Using DevOps for AI will increase the velocity of model development and delivery while simultaneously improving the quality of it. This will be done by reducing non value-add tasks and accelerating data preprocessing.
Where to start
While the concept of combining DevOps and AI is not new, there are few companies that have implemented this. Part of the reason for this is that companies might know DevOps or they might know AI, but they don't know both. While DevOps has become mainstream, AI has historically been something that companies have dabbled with, either by creating some limited POCs or identifying a use case or two where AI could be of benefit. While both of these methods are valid, they come with steep learning curves, are slow to develop and don't necessarily provide the expected results. At the speed that the marketplace and technologies are changing these days, we recommend that you choose a partner to help you jumpstart your DevOps and AI journey.
If you're truly serious about AIOps or DevOps for AI, then you need someone who is experienced with both DevOps and AI. Endava has a long history of delivering projects using DevOps and an almost equally long history in building and deploying AI models. If you combine our experience with your business, the result could be something truly fantastic.
Robert Anderson is principle insurance architect at Endava
