DATAx New York: How financial services build in-house ML teams

In a data panel comprising data experts from financial institutions, among the many topics covered, we focus on a conversation on whether to look internally or externally for your ML needs

12Dec

This afternoon's banking panel moderated by Eric Thornton, featuring Martin Caupin, a data scientist at BNP Paribas and Devanshu Mehrotra, vice-president of audit data analytics at MUFG, the panel delved into how data analytics is currently driving the future of banking.

The panel engaged the audience in new innovations such as the move toward cognitive automation, the approach of viewing data as you would with any other physical assets with a lifecycle, and the question of who you attribute ownership of data you have used to build ML models in-house.

However, one of the biggest topics brought up was the fundamental issue firms face trying to create their own ML engines. As Caupin explained: "Something I have seen with companies trying to build ML models is that they bring in external experts to build "something" in-house (because they aren't sure what) and they find that they aren't rich enough to maintain them and the experts eventually get bored and, in the end, they just leave.

"Few industries have what it takes to keep these experts on, financially and interest-wise; it is mostly just tech companies and very large financial institutions. To utilize these experts effectively, you need to make sure they are fed the right data and they know exactly what is expected of them," he added.

However, these experts are simply too few and too in demand, and while it may be cost effective to bring them in to help build the essential models which are needed today, in order to build a long-term data strategy, Mehrotra said you need to start creating interest internally.

"Please don't groan when you say this, but one way we have started to create interest within our firm is by releasing a data analytics newsletter. It is a video newsletter; I find it odd that the way we interact with media at work is so different from the way we do it at home. We watch multimedia content from a dozen different channels, but we get to work and step 20 years into the past, and we're reading everything…if we are pushing innovation, we also need to be engaging and innovating in-house.

"When you enter the ML field, you are not just competing against other financial institutions, you are also competing with tech giants such as Google and Amazon. Therefore, if we start educating in-house at the analyst level, with newsletters or with programs like

Champion or mentor programs, you are on track to building your own ML teams."

And this not only benefits the company long term but impacts the individual greatly.

"Employees know as well as you do that they need to find added benefit to their roles. If you're an auditor and you are still processing information the old school way…you're not going to have a job 10 years from now and I'm saying that as a former auditor," said Mehrotra.

"You need to keep yourself relevant and these programs give people the ability to do exactly that."


Other experts from the financial sector spoke this afternoon such as Victor Tewari from BMO Financial Group presentation on how to drive holistic data to drive business decisions with machine learning and data analytics along with other talks from Chameleon Metadata and JP Morgan.

Click here to learn more

How big data can protect infrastructure small

Read next:

How big data can protect infrastructure

i