Pete: How do you overcome the complexity and greater intelligence, but still avoid breaking the bank? Well, it starts with lowering the barrier to adoption everywhere that computing takes place. Intel is at the heart of standardizing and proliferating technology for over 50 years. And now we’re at it again with artificial intelligence, with a wide selection of productivity tools that span a line-up of general purpose to domain specific technologies, all built on common standards. We at Intel like to make it easier, faster and more cost-effective for everyone to infuse AI and analytics into apps from Edge, to cloud, to client and unlocking and valuable insights in the process.
But don’t take my word for it. Let’s learn from our guest, Sebastian from Quantifi about how they’re using our products from Intel to delight their customers. So, welcome and I’ll ask you to introduce yourself.
Sebastian: My name is Sebastian Hahn and I’m the Artificial Intelligence and machine learning Lead for Quantifi. Quantifi is a FinTech provider to global banks, investment managers, and corporations, and we help clients to leverage new technology to better trade and manage risk.
Pete: Terrific, great choice for our discussion today. So let’s start with the first question, just to help us understand a little bit of on who your end users are and really how they’re using this technology?
Sebastian: Quantifi’s clients are buy-side and sell-side institutions, including hedge funds, pension funds, sovereign wealth funds, and banks. We also work with professional services firms and technology companies. Typically end-users within those firms would be traders, portfolio managers and risk managers.
One of the main challenges that we help these institutions with is calculating valuations and risk metrics, such as sensitivities or stress tests, on their portfolios and they want to do that in a reasonable timeframe. For some calculations, the computations can be so complex that it can take a long time and it becomes challenging to manage one’s risk if the necessary risk data is not available quickly enough.
Pete: Tell us a bit about Quantifi and give us a little context on the innovations that you’re making when it comes to solving some of those problems you discussed, for your users. And then where your collaboration with Intel comes into play?
Sebastian: Sure, the portfolio management and risk systems that we provide for to buy-side and sell-side firms are often front-to-risk systems that cover many different aspects of the organization. However, at the core of these offerings are our own proprietary models that help clients to value derivatives, as well as other products, and calculate risk metrics on them.
Now, the challenge that I mentioned earlier is that these complex portfolios can sometimes be hard to value and hard to analyze. This is where we are working together with Intel on improving the performance of these models and achieving faster computations for our clients, so that they really can see their risk, as close as possible to real-time.
There two ways in which we’ve recently worked with Intel that I would like to highlight. The first instance would be for accelerating derivatives valuations using AI and machine learning techniques. By leveraging Intel’s newest Ice Lake processors we have achieved valuation speed-ups of 700x. We have also been working with Intel on accelerating XVA performance using Intel’s 2nd Gen CPUs and Intel Optane Persistent Memory to improve IO and total performance.
Both of these innovations empower our clients to conduct in-depth risk analysis more effectively.
Pete: I love to hear stories of 700x speed ups. Just help the audience a little bit more, understand about how you’re working with us together to leverage our AI portfolio of products and accelerators to delight your customers.
Sebastian: As I mentioned, one of the challenges that our end users are facing is being able to calculate valuation and risk metrics within a reasonable timeframe and many financial institutions actually don’t know their risk in real-time. Working with Intel, we’ve taken some conventional finance models, based on numerical integration and Monte-Carlo techniques, and trained an artificial neural network on one of those models and basically taught the network how to imitate the behaviour of our conventional proprietary model.
We then worked with Intel engineers at optimizing the performance of that neural network in terms of speed and accuracy. Using Intel’s OpenVINO Toolkit, and also the Ice Lake processors we achieved a speed up of 700x, which really can make a massive difference in the life of the portfolio manager or a trader bringing calculations down from hours to seconds.
Pete: That’s fantastic. This was just a fascinating conversation on really some cutting edge work. It’s a pleasure to hear about just the different ways you’re working to deliver your products and solutions along with Xeon scalable based platforms together, right alongside you to deliver solutions that are going to work your customers and ours.