At the upcoming STAC Summit in London, AI in asset management is front and centre with a presentation from Gaurav Chakravorty, co-founder of hedge fund Qplum.
Chakravorty will be talking about AI from the CIO’s office to the trading desk, focusing on collaborative filtering for investments and deep learning for trades.
The hedge fund has been researching different approaches to longer-term investment strategies, in particular how to fine-tune allocations to strategies. One approach is collaborative filtering, a technique similar to recommendations systems like Amazon’s.
Qplum will also be presenting at New York’s STAC Summit, with machine learning engineer Hardik Patel discussing what makes machine learning in finance difficult.
Peter Lankford, STAC’s director, explained that machine learning in finance has its own idiosyncrasies compared to other industries and application areas, so Patel will be able to provide an inside look at the ways that is the case, and how to deal with it.
Also in New York, $50 billion quantitative firm Two Sigma will talk about a new way of improving both performance and usability of Python in Spark, which is one of the more popular approaches to data science used by AI developers.
Data science has been a major investment for hedge funds, and it’s resulted in some do-overs as early movers adapt to the reality of what can be accomplished versus the hype.
Bloomberg reported recently that Point72 has struggled along with other hedge funds in turning big data into big profits and, as a result, chairman and CEO Steven Cohen revamped the big data operations, again.
It’s not an isolated incident either.
At London’s STAC Summit, Michel Debiche, a former quant on both the buy and sell sides now turned consultant, will be talking about why the process of data science doesn’t work as well as it could in many cases.
“It’s happening in a number of firms. Both hedge funds and big asset managers are under pressure to up their game in terms of their ability to get more use out of the data they have and as well as use new sources of data,” said Lankford.
“And this is causing a rethink about both about technology and process work in data science, and Michel is going to talk about how a more industrialized process around data science makes sense,” he added.
One of the biggest complaints about data is on the topic of quality control, it’s also one of the biggest problems experts warn of for anyone diving into AI algo development.
“It’s about doing things at scale,” explained Lankford. “Data scientists and analysts, typically they are very independent and take different approaches, and there’s an aspect to coordination that comes into play as well as quality control.”
Hardware and cloud
FPGAs have stayed on the radar because of how applicable hardware acceleration is for AI, among other analytic workloads. Moreover, the chips are very power efficient.
Sounds great, but challenges remain in terms of programming in low level languages, so STAC will be providing a “pulse check” on moves to make higher level languages accessible to developers.
Also relevant to AI, a look at cloud platform developments. Nearly all fintech and regtech firms have built natively in the cloud, with many decisions based on low cost and responsive infrastructure. As that evolved, analytics and machine learning took off on the back of such developments.
“It’s certainly an obvious starting point for a lot of data scientists, they can quickly spin up resources and see how things behave,” said Lankford. “Whether they stay in the cloud or not is another question, but there is a lot of momentum.”
Still, there are also many issues, questions, and obstacles, so STAC has created a special interest group on cloud that includes banks and hedge funds primarily.
“We’re facilitating dialogue and conducting testing around these issues and some of them have a bearing on AI,” Lankford said, “That’s going to be a quickly growing area of research for us.”
The London STAC Summit takes place April 26, with Chicago following on May 7 and New York on June 13. Follow the events.