Startup AI firm impresses with massively crowd sourced biometrics algorithm training

DefinedCrowd, which bills itself as a crowd-as-a-service AI firm, has closed a $50.5 million series B funding round destined for product development and the expansion of its sales and support network.

The company’s products include training for natural language processing, speech synthesis and computer vision, such as helping make the facial recognition technology of a Fortune 500 global electronics maker more inclusive, according to a company website post. DefinedCrowd’s clients can be found in retail, fintech, automotive, energy, health care and media.

Executives say that among their customers are BMW AG and Mastercard Inc., both of which have much to gain from accurate, broad-based voice interfaces.

A post on the company’s site reports that revenue has grown 656 percent year over year, and staffing has increased 176 percent over the same period.

The 4.5-year-old startup creates data sets that are built by 290,000 paid human contributors speaking 50 languages and by machine learning, according to GeekWire.

That community could be a significant advantage for DefinedCrowd. Rather than collecting data from their clients’ customers (and thereby almost inevitably inviting complaints from privacy advocates), the company employs its crowd to transcribe information, record their own voices and other activities used to train algorithms. Human-in-the-loop annotation procedures and quality assurance with machine learning are used to ensure precision in the training process.

Amazon and Sony are among DefinedCrowd’s longer-term investors using their corporate venture capital arms, the Alexa Fund and the Innovation Fund, respectively. Mastercard also has been onboard for previous rounds.

Other investors include Kibo Ventures, Hermes GPE, EDP Ventures, Semapa Next, Evolution Equity Partners, Bynd Venture Capital, Portugal Ventures and IronFire Ventures.

Speaking with Bloomberg, the company’s chief executive, Daniela Braga, is quoted as saying she is on “the road to an IPO,” probably within five years.

Article Topics

AI  |  artificial intelligence  |  biometrics  |  computer vision  |  dataset  |  facial recognition  |  funding  |  machine learning  |  speech recognition  |  training