When we spoke to Sameer Gupta, a former COO for J.P. Morgan’s global electronic equities trading business, last month, he said banks have been hiring all the wrong people for machine learning and artificial intelligence (AI) divisions. Banks’ teams have a tendency to over-complicate things and achieve nothing, said Gupta. This might be why J.P. Morgan has recruited a new global head of machine learning from outside of the industry.
Geoffrey Zweig joined J.P. Morgan in New York City last month. A Berkeley PhD specializing in natural language processing (NLP), Zweig has literally zero background in finance. Instead, he’s spent his entire career working for either Microsoft or IBM. In his most recent role at Microsoft, he developed a, ‘speech recognizer to achieve human parity in conversational speech recognition. ‘ The new product caused excitement last year when Microsoft’s chief speech scientist, Xuedong Huang, wrote on the company blog that the company had “made a major breakthrough” in speech recognition thanks to Zweig’s team. “We were able to move more quickly than we anticipated,” Zweig told Business Insider.
Snaring one of the world’s foremost NLP specialists for what looks like a new machine learning division looks like a big win for J.P. Morgan, However, the bank has been mute about his arrival. Zweig is quietly tasked with using J.P. Morgan’s data-sets to develop a machine learning strategy that can be rolled out across the firm.
This isn’t J.P. Morgan’s first foray into machine learning. Graham Giller, its former chief data scientist (hired from Bloomberg in 2015) moved into a role encompassing advanced analytics, data science and machine learning in January 2016. Following Zweig’s arrival, Giller has become head of data science research for the corporate and investment bank (CIB). The chief data science officer for J.P. Morgan’s CIB is Afsheen Afshar, a Stanford graduate hired from Goldman, who also arrived in January 2016.
Afshar spent last year building J.P. Morgan’s data science team. Nonetheless, the bank seems light on NLP talent. It seems likely, therefore, that Zweig will be hiring as he develops the bank’s new product. His former team at Microsoft, which he describes as “world class”, looks like the most likely target. As Gupta’s colleague noted: “It’s extraordinarily difficult to find people who can build NLP systems.” – Much harder than finding people who know about finance.