JP Morgan Asset Management’s Lee Bray Discusses Data Science


Following his appointment in July as J.P. Morgan Asset Management’s APAC Head of Data Science, Lee Bray has been busy building a dedicated team in the region to support the firm’s ongoing mission to strengthen its investment platform and accelerate innovations in artificial intelligence and natural language processing.

Bray was the main speaker during a session on Data Science in Trading at FIX’s 19th Asia Pacific Trading Summit, held 2 September. Among his messages to the audience was the need for data scientists to be good listeners in gauging the most pressing problems across business lines.

“A good data scientist is someone who can work closely and collaboratively with people who are probably more experienced in their subject matter, then translate that into a data science outcome,” Bray explained. “It is not easy to get a data scientist who is as proficient at the collaborative part as they are in the technical aspects of the job.”

Bray shared several other insights on his firm’s data science journey in response to the following questions:

Lee Bray, JPMAM

Please tell us generally about some of the ways J.P. Morgan is using machine learning.

Trading in general is moving more towards looking at using data science in aspects outside the low touch world, but it’s early days. My hope is we move there quicker rather than slower.

We adopted ML [machine learning] techniques about 3 or 4 years ago on the desk, predominantly on the low-touch side. We’ve expanded that to incorporate IOIs [Indications of Interest] and have put a more systematic approach around how traders interact with brokers. We’ve developed techniques which will recommend what brokers to call with our larger size orders in the hope that they will have an increased probability of finding us a block.

Another thing that we really put a lot of time into, which has taken a significant amount of effort, is trying to come up with a solution to high-touch trade broker selection, we now give our traders recommendations on what algos and brokers to use. This is a significant step forward in my view.

How are you using data science beyond trading?

We are also looking at it from a creation of funds perspective. J.P. Morgan Asset Management has a number of funds that are specifically focused on using data science techniques. We have a proprietary natural language processing tool called ThemeBot, which combines big data research and artificial intelligence, it creates a data science approach to investment.

One of the largest user cases is creating efficiencies. One example we have currently is an RFP tool. This is a process that a lot of firms will have significant resources allocated to, and is almost always manual i.e. filling out answers to questions manually. When you use natural language processing techniques it is possible to automatically match questions with answers and only afterwards have someone check those outcomes. That makes it significantly more efficient operationally. Although that might not seem as exciting as talking about trading, those user cases across a big firm like J.P. Morgan can make a large difference in terms of improving efficiency.

Another one worth mentioning is around the use of data science on the risk side. We’ve developed a crowded trade chart, used by our risk department, which takes in market data and also public filings to identify any areas of crowded trades.

Can data infrastructure and accessibility pose a challenge?

Anyone in trading will probably appreciate that data in in this area is quite abundant, and in my view having looked at it more holistically, quite easily accessible. The barriers to entry for using data science techniques on the trading desk are relatively low. For example, just as an initial step, people can access tick databases to start using these techniques.

For the rest of the firm – and I would say this is probably the same for all large firms – data infrastructure is not easy because many of the systems that are outside of the trading world were probably created prior to data science even being a term.

As part of my new role as APAC Head of Data Science, my team will partner with internal teams that have been tasked with making sure that data is accessible into the future. In this way we can work towards future proofing data access.

Over the past year, the markets have made rigorous demands on data science. When will it be the other way around with data science leading the markets?

Not for a while. I think it’s complementary, particularly in trading and fund management. We support investors in managing their money, there’s always going to be a considerable human overlay around the investment process. I simply cannot see that changing any time soon.