Mark Sheldon has worked in software, machine learning and artificial intelligence for almost his entire career, and so has a high-level and long-term view of the industry.
Therefore, it was only right to ask him if the phrase ‘artificial intelligence’ is as overhyped as some people say it is. His response: it may be more of a case of misunderstanding than deliberate misuse.
He said: “Many business leaders don’t fully appreciate the breadth of what AI actually represents from the traditional machine learning and statistics that have been around for a long time.” He added that while telcos have been using statistics and machine learning for over 20 years, the newer fields of robotics and image processing are getting many more headlines and there is probably just a misunderstanding of what AI represents because it is such a broad field.
Sheldon has just been appointed CTO of Sidetrade, having previously led their AI sales and marketing platform, integrating that platform into the wider technology stack of the company. He had founded a predictive analytics company, BrightTarget, in 2013 which was acquired by Sidetrade in 2016.
Now he is responsible for rolling out the use of AI across Sidetrade’s portfolio which comprises two main products; one that automates processes around cash collection and another that optimises the journey from customer acquisition to the growth in value of those customers.
“The whole scope of what we do is accelerating customers from the acquisition to the collection of cash,” he said.
In this context, I wondered where the data professionals sit within this organisation as well as how they interacted with the wider business. Sheldon said that he has approximately 80 data professionals organised in small agile teams or squads made up of software engineers, data engineers, devops and data scientists.
“We have influence in the organisation where we mix traditional data scientists among the R&D structure in a very agile way. It is fairly innovative but I see it more and more now as best practice across the industry,” he said.
Sheldon said it was a strategic decision to situate the data science team as close as possible to R&D. This way the data professionals can push innovations to customers so they learn and give feedback to the data science team which benefits from agile iterative behaviour.
One thing that he and his team have been working on is a way to reduce the amount of time data scientists spend preparing data and getting it ready to run models on it. To that end, he wants to automate what he calls ‘commodity data science’ which would involve the automation of data processing, the regular execution of models and of different testing.
“We spend a lot of time in what I call the ‘AI ops’ world which is a trend that has started to emerge in this space over the past two or three years. It has come from how do we actually productionise all of the modelling and the stuff the data scientists are producing, how do we run it in production and how do we monitor it and support it going forward.”
The biggest challenge Sheldon sees when it comes to productionising data science models is simply the volume he and his team have to deal with. “Across our software platforms we train thousands of different predictive models for different clients and they are changing all of the time in terms of the data they are trained on. So actually building workflows and processes to help monitor those in production at that kind of scale without having to scale the team in proportion is probably the biggest challenge.”