Yandex Turns Its Machine Learning Smarts Into A New B2B Product, Yandex Data Factory

Russian search giant Yandex — which has just made an acquisition to expand its e-commerce operations — is also unveiling a new product that will expand its deeper into the world of B2B services. Today it is launching Yandex Data Factory, a new business unit that is based on the machine learning that it has developed internally for its own services, called MatrixNet, which has now been developed into its own product.

YDF will be based around an SaaS, subscription business-model and the company says that there are some 20 projects already being developed for clients.

YDF is part of publicly-traded Yandex’s bigger ambition also to grow its business beyond its largely Russian boundaries and its core search-ads business. The unit will have a dual HQ in Amsterdam and Moscow, and tellingly, the service is getting officially unveiled not in Russia but in Paris at the le Web conference currently underway.

YDF is not being spun out of thin air. The company has been working on a one-to-one basis with organisations like banks, Russia’s road management agency and others since earlier this year to apply its machine learning technology and artificial intelligence to different challenges and tasks for these third parties — effectively using the same technology that the company already uses for its own search, music recommendations and speech and image recognition.

This included a collaboration with CERN, in which Yandex “trained” MatrixNet to search for specific particle collisions among events registered by the Large Hadron Collider beauty (LHCb).

“The success of this project gave us reasons to believe it can be repeated in other areas of application,” the company writes in a blog post. “Any industry producing large amounts of data and focused on business goals could benefit from our expertise and our MatrixNet-based technologies: personalisation of search suggestions, recommendations or search results, image or speech recognition, road traffic monitoring and prediction, word form prediction and ranking for automated translation, demographic profiling for audience targeting.”

But actually the work goes back even further than this year. In 2012, Yandex also took a stake in a geo-tech business, Seismotech, as part of its strategy to apply its computing architecture and algorithms into solving other problems.

YDF is not unlike what Google is developing in its own range of enterprise services and through projects like Google Brain (and what ex-Googlers who have left the company are also trying to develop).

Today, companies like Yandex and Google make the bulk of their revenues from search ads, but to diversify, and to offset whatever declines that business may see in the future, they need to continue developing and investing in new areas like this, and so we’re likely to see more of this to come.