Designing, training, and validating sophisticated machine learning algorithms can suck up a lot of hours — particularly if you’re dealing with sparse datasets. This has given rise to a cottage industry of crowdsourced data science, led by Google’s Kaggle and companies like Topcoder. And the market’s growth potential is getting investors to sit up and take notice.
Sunnyvale, California-based CrowdAnalytix, which describes itself as a “library” of data-extracting bots built through competition, is the latest to throw its hat in the ring. The six-year-old startup today announced that it has raised $40 million in a strategic investment from Macnica, a subsidiary of Tokyo electronics firm Macnica Fuji Electronics Holdings. The investment will be used to buy out shares from existing shareholders as well as infuse capital into CrowdAnalytix.
This brings the startup’s total raised to $43 million, following a $1 million round in December 2016 and a $2 million round in May 2012.
CrowdAnalytix CEO Divyabh Mishra said the capital will be used to expand the company’s reach into the broader Japanese market and new verticals where Macnica has a strong presence, like manufacturing and health care.
“Macnica’s strategic investment validates the huge market potential of our industry-leading AI platform,” Mishra said “We look forward to deepening our synergy with Macnica in cultivating joint customers, developing complementary technology offerings, and expanding globally together.”
CrowdAnalytix’ platform consists of two layers: a machine layer and a human layer. The former comprises an API-accessible suite of over 500 artificial intelligence (AI) algorithms that collectively extract more than 2 million attributes per hour from text, images, and audio. As for the human layer, it’s a community of over 20,000 data scientists who compete for cash prizes to create algorithms that can, for example, identify superheroes from product images or forecast the bidding price of generic drugs.
The winning algorithms migrate to CrowdAnalytix’ Datax portal, where they’re continuously monitored for dips in precision. Any that degrade below a certain threshold migrate back to the community portal for retraining or rebuilding.
Among the AI systems the community has helped create is an attribute extractor that classifies products into one of 5,500 types, a price-matching ecommerce tool that finds and matches the price of items on competitors’ sites, and a descriptive content creator that generates product descriptions and titles from attributes.
CrowdAnalytix says retail clients using its automatic classifiers — which include Fortune 100 companies — see an average 35 percent reduction in product returns, a 20 times decrease in cart abandonment (i.e., when an online shopper who has added items to their cart leaves the website without purchasing anything), and a 90 percent increase in onboarding.
For Macnica, a decades-old company that ships semiconductors, network equipment, and software in the electronics and IT markets, the investment in CrowdAnalytix — and its community of data scientists, by extension — offers an opportunity to expand its AI business.
“Artificial intelligence is transforming every industry as we know it, and CrowdAnalytix stands out as a clear leader driving the underlying technology innovations,” said Macnica president Kiyoshi Nakashima. “We look forward to incorporating CrowdAnalytix’ AI solutions into our enterprise offerings and to helping our clients gain sharper competitive advantages in the marketplace.”
CrowdAnalytix’ previous investors include Accel Partners and SAIF Partners.