The Commonwealth Bank of Australia is laying the groundwork to train its own Kaggle 'grandmasters’, as it readies to meet a 12-month deadline.
At an H2O World Sydney event in November last year, CBA's chief executive Matt Comyn first expressed that he was “extremely hopeful” that "over the next year or two" the bank would be home to some of its own ‘grandmasters’.
Grandmaster is the term the Kaggle data science community uses to describe the top-ranked participants; there are currently 300-plus worldwide, and they are seen in some circles as a pinnacle of data science talent.
CBA is hoping to build its own grandmaster team, partly through its investment in H2O.ai, a cloud-based machine learning platform that helps customers build and scale AI-based services.
Experience and support
Speaking with iTnews, CBA's chief data and analytics officer Dr Andrew McMullan said the bank is now laying the groundwork to reach its grandmaster goal.
“We're giving our data science community the practical skills and experience and support to be better at the things that they do," McMullan said.
“Then we will encourage them through a series of internal and external events to apply that craft to the international competition environment so that they can demonstrate how good they are.”
McMullan said that “there's absolutely no reason why Australia can't be seen as a leading nation” in AI capabilities.
CBA is playing its part by building internal capability and working with “academic institutions and governments on what can we do to help Australia be seen as a leading nation in AI”.
With Comyn setting “challenging targets”, McMullan said he was “very confident” that CBA could build a capability that is "absolutely as good as" what it had been able to source from its H2O.ai partnership.
One of the reasons the bank struck the partnership was to gain access to the grandmasters that H2O.ai employs.
“When you work with incredible talent, it lifts the expectations that you have of yourself, and I think it helps just raise the capability," he said.
“That's exactly why we wanted to partner with H2O.ai, and we'll continue to partner with them.
"But we absolutely have an aspiration to build world-class capability at CBA, invest in our people to be as good as anyone else in the world.”
He said CBA has “an extremely capable data science team today”, with its latest influx of graduates a testament to the bank's “commitment to the talent pool across Australia”.
With the new graduates, which include CBA’s largest cohort of data scientists, surrounded by the right support system, McMullan said it may be that at least one of them “becomes a Kaggle grandmaster.”
Grandmasters may also be found among existing staff as well: the bank is “investing in our people” through reskilling programs and working with staff who express interest in AI and machine learning analytics, McMullan said.
CBA’s overall “aspiration is to see how quickly we can get our first” grandmaster and develop a cohort from there, with its main target “to be an AI-first customer-obsessed organisation.”
“We're definitely investing in a different capability of skillsets in the workforce and that's why we're so focused here because we know that if people use a lot more advanced techniques … we can make better decisions to serve our customers every single time we interact with them."
McMullan also said the pathway to becoming a grandmaster “takes a little bit of time” as participants need to win at least five gold medals in competitions.
He said CBA works directly with H2O.ai Kaggle grandmasters including triple grandmaster, Shivam Bansal.
Working with Bansal means CBA can tap into his knowledge to assist with improving understanding customer feedback and leveraging natural language processing tools.
New graduates
McMullan said that alongside aspirations its new data scientist graduates will step into the pathway of its grandmaster goals, he found overall new students “unbelievably qualified to be able to use advanced techniques that they've learned at universities”.
“The courses today are very much practical application of data science techniques," he said.
"We've got people who've done natural language processing, neural networks, all of the deep learning capabilities.
"When people come to us today the skills and experience and expertise that they have is so far beyond what we would have got maybe three-to-five years ago.”
