What do you get when you mix an Iranian born-and-educated CEO, two college dropouts, a stint in China, a short-lived consumer business and the Irish startup visa program? A natural-language processing startup called Aylien, naturally. Although, jokes aside, it might represent something much larger than the company’s business acumen if Aylien is able to succeed in this field.
Aylien, which launched in 2010 as a consumer business with a product called Talk.ee (“basically, a better Reddit,” founder and CEO Parsa Ghaffari said), pivoted in late February into the business-to-business world with a collection of API services that expose its text-analysis capabilities to interested developers. “[We] realized that maybe we are not a consumer team by nature,” Ghaffari said. “It’s like going to the gym because the cool guys are doing it.”
Among the new products is a text-analysis API that performs tasks such as classification, entity and concept extraction, language recognition and summarization. It’s also working on a few others including an API for building intelligent news-reader apps, and some for social-media management and news tickers for websites. As of last week, Aylien claimed about 40 paying users, Ghaffari said, largely in the publishing, social media and legal industries (i.e., ones that can benefit from automatically classifying and tagging content).
However, as interesting as the company’s history might be (it has pretty much moved from one startup haven to another since launching in 2010) and even as good as its products might be, Aylien might be most interesting because it’s not staffed by a team of machine-learning experts from top universities, as is so often the case with companies or projects tackling natural-language processing. No, Aylien is a bunch of programmers who learned some NLP to build Talk.ee and then taught themselves even more when they decided to start selling NLP as a service.
“We are not really here to improve natural-language processing or machine learning,” Ghaffari said. “We’re here to make them mainstream.” He thinks the company’s focus on real-world applications will help it make money and open the technology to more users than would be possible if it were constantly researching the next best techniques.
It’s an extreme take on one of the big themes of our Structure Data conference this year (just two weeks away now, March 19-20 in New York), which is that advanced machine learning and artifical intelligence capabilities are getting much easier for mere mortals to consume. We have a talk dedicated to this idea with Eliot Turner of AlchemyAPI (almost certainly the upstart Aylien’s biggest competition) and Stephen Gold of IBM’s Watson group, but several other speakers are aiming to do similar things. Size aside, the big difference between Aylien and these companies is that they still rely on engineering and even leadership teams full of machine learning experts and Ph.Ds.
Aylien is a product of an even newer world, one where there’s enough free information floating around that a group of less-pedigreed developers can learn enough to start a business around NLP that might well be good enough to solve some real problems for businesses. I don’t want to oversell the importance of Aylien or its NLP prowess, but if it’s able to succeed in an NLP space dominated by Stanford and MIT alumni, that might say a lot about how much faster we can expect new technologies to hit the mainstream in the coming years, especially when they have the cloud to deliver them.