First Derivatives’ Kx Goes Deep on Machine Learning

The first prong of Kx’s strategy is a consultancy service. Kx—which offers in-memory computing and streaming analytics through its kdb+ time-series database—has created a team focused on machine learning in London, led by Mark Sykes.

The team is currently smaller than 10 people, but Sykes tells WatersTechnology that it is actively adding staff. Additionally, the firm has partnered with machine-learning consultancy Brainpool, which has 130 machine-learning engineers across commercial and academic institutions, while Kx’s second area of focus is on software development.

“What we intend to do between now and the end of the year is create open-source, free machine-learning libraries that allow kdb+ to natively implement machine-learning algorithms or integrate into third-party tools that are in use by our customers—such as Google’s TensorFlow, the Keras neural network library, and the Microsoft Cognitive Toolkit package—to allow those to reside within the Kx environment and infrastructure,” Sykes says.

Sykes confirms that two customers have already asked Kx to “initiate active projects…to apply this library and advise them on how to best implement systems with specific business objectives,” although he was not able to provide greater detail about the projects due to NDAs.

In addition to its kdb+ database and its proprietary Q array processing language, Kx also offers a surveillance product, a transaction-cost analysis (TCA) tool, an FX analytics product, and a service to help clients prepare for the Consolidated Audit Trail’s (CAT’s) implementation. Sykes says that in the future, Kx will look to deploy machine-learning techniques within those offerings.

“There’s a third element we will introduce once we’ve got the machine-learning libraries into their initial release state, and that will be where we use those libraries to apply machine learning to our own set of solutions,” Sykes explains. “We are aiming to bring new levels of functionality to those solutions by applying machine learning ourselves to our own products. So we will be re-releasing add-ons to those using our own machine-learning libraries beginning next year.”