Ask the expert: considerations in AI deployments

In this week’s Ask the Expert segment, Romain Bottier, senior solutions architect for high performance computing and artificial intelligence (AI) at Dell Technologies South Asia, advises organisations on what to consider in their AI strategy.

Q: I’m currently using some AI features in business applications from vendors like, but I’d like to go one step further and leverage AI to gain a competitive advantage. Could you provide some advice on “build vs buy” when it comes to AI implementations? Perhaps a framework I can follow to help with the decision?

Bottier: This is a good question and a trend that we have seen emerging. Cloud platforms have played an important role in democratising the use of machine learning (ML) for enterprises by simplifying and disaggregating hardware and software stacks. With the help of specialised cloud providers and public cloud providers, many enterprises have implemented ML and some form of deep learning (DL) in a supervised environment, without having to make significant changes to their IT infrastructure. The market and independent software vendor (ISV) ecosystem around ML and DL has increased considerably with a plethora of options available, from cloud to on-premises or ultimately hybrid cloud environments.

When you are looking at applying AI in business, there are a few key elements you must look at:

  • Data: what, where and velocity? It is important to identify what is the data you have, where it is located and the volume as well as the flow.
  • Compute: how much training capability do I need?
  • Platform: any platform, frameworks, programming language I am familiar with?
  • Production: How do you transition to the production use of ML/DL at scale?
  • Total cost of ownership: opex or capex, which one suits you the best?

There isn’t any magical combination that fits everybody’s needs, but one of the most widely used environment when it comes to AI so far, is the hybrid cloud design. In adopting a hybrid cloud, you can benefit from the best of both worlds: choice of platforms and tools to implement end-to-end AI at scale, without concerns of IT complexity.

As an example, you can build a predictive model based on your business data from your preferred CRM platform. The insights generated can then be combined with other datasets that aren’t available within the CRM platform to create another model using AutoML. This model can be part of an app containerised and deployed at scale within your company or on any cloud for your customers.

Data Center
Data Management