In November, Riskfuel was selected to participate in the annual Innovators Pavilion at the FIA Expo. The FIA Innovators Pavilion showcases fintech startups to encourage greater innovation in the derivatives industry. The startups were evaluated on the degree of innovation, the relevance to derivatives markets, and the potential impact on the industry and judges chose Riskfuel for “its innovative solution to the challenges of derivatives valuation”.
The following is a transcript of the winning pitch presented by Ryan Ferguson, CEO and founder.
For 15 years I worked as a derivatives trader – trading both cleared and OTC derivatives across a variety of asset classes including credit, equities and rates. Over those years, technological change improved every bit of the trade lifecycle of my cleared trades. But the OTC side did not keep pace.
The central issue is the massive compute bottleneck associated with valuing and calculating risk sensitivities. Everything is calculated after the markets closes in a big, expensive nightly batch. Everything. the P/L, the P/L explain, the Risk reports. The quality isn’t great either. To keep the costs under control the models are often dialled down to just acceptable levels.
Riskfuel uses deep learning/AI to dramatically accelerate the slow models, breaking the bottleneck and allowing for low cost generation of timely and accurate valuations and risk sensitivities. Riskfuel accelerates models so that they are millions of times faster. Earlier this year, we collaborated on a study with Microsoft to accelerate a client’s FX barrier option model. This was a real model used in production with all the bells and whistles. Not a prototype or an experiment. The client model produces 35 valuations per second. The Riskfuel version of this model reached nearly 1 billion valuations per second.
Crucially, Riskfuel models are accurate. The output produced is indistinguishable from that of the target model.
Who will benefit from this innovation? Traders with OTC positions will be confident, real-time participants in the markets. Real-time market data will drive real-time P/L reporting, explain and risk sensitivity calculations.
Their managers will have an accurate picture of performance and risk at every moment throughout the trading day, aggregated across all of their business lines, both cleared and OTC. And management will definitely be pleased by the huge savings in compute costs. The broader public will also benefit with i) more efficient OTC markets will lead to better pricing for end users, ii) inexpensive stress-testing will improve the safety of our institutions, and iii) lower data center utilization will help companies reduce their environmental footprint.
How difficult is implementation? Attaining these benefits is within reach. Implementation of Riskfuel models is a low-risk proposition. We can accelerate a single problematic model or a whole host of models. At the same time or incrementally. The Riskfuel accelerated model takes the same inputs and produces the same outputs as the existing model. There is no expensive plumbing that needs to be done. You can fully leverage your existing platform.
Let’s get started. We’ve got work to do.