Sherpa, A startup from Bilbao, Spain, which was a voice-based digital assistant and an early proponent in building predictive search for Spanish-speaking audiences, has raised some more money for the startup to double up on a new focus : The creation of privacy-first AI services for enterprise customers.
The company has closed down $ 8.5 million, noting that Sherby founder and CEO Zaibi Uribe-Etekaberia said it would be used to continue building a privacy-focused machine learning platform based on an interactive learning model , Which accompanies the existing conversational AI and search services. Early users of the service include Spanish public health services, which were Using the platform to analyze information on COVID-19 cases Predicting demand and capacity in emergency rooms across the country.
Apex Digital has funding coming from managing partner Marcello Gigliani; Alex Cruz, President of British Airways; And Spanish investment firms Mundi Ventures and EckerPen. The funding is an extension of the $ 15 million that Sherpa has already raised Series a. From what I understand, Sherpa is currently picking up a big series B as well.
The creation and commercialization of joint learning services comes at a time when conversational AI business has found itself stalled.
Sherpa saw some early traction for its Spanish voice assistant, which first emerged at a time when Siri as Alexa, Apple’s efforts as Amazon, and others to address markets outside of those in fact Did not make strong advances where English is spoken.
The service passed 5 million users by 2019 – customers using its conversational AI and predictive search services include Spanish media company Prisa, Volkswagen, Porsche and Samsung.
But as Uribe-Etxebarria describes it, while the supporting business is still together, he came up against a difficult truth: the biggest players in English voice assistants eventually added Spanish, and the AI they would invest over time. Yes, he will make it. It is impossible for Sherpa to survive in that market for a long time on its own.
“We won’t be able to compete against Amazon, Apple and others until we make a big deal with a company,” he said.
Due to which the company started searching for other ways to deploy its AI engine.
It came to condensed secrecy, Uribe-Atkebaria said, when he began to see how he could expand his predictive search services into productivity applications.
“An ideal assistant can read the email and know what action to take, but there are privacy issues about how to do that work,” Uribe-Etexebaria said. Someone suggested to her that she learn a way to work with the federation by emailing her assistant. “We thought, if we ask 20 people to do the work, we can create something to read emails and respond to.”
The platform Sherpa built, Uribe-Etekberia said, did a better job than she had anticipated, and so a year later, the team decided she could use it more for email only: product it Can be built and sold to others as engines in a more privacy-compatible way to model training machine learning with more sensitive data.
It is not the only company pursuing this approach: TensorFlow from Google also uses federated learning, as it does fate (Which includes cloud computing security experts from Tencent contributions), and PySyft, A joint learning open-source library.
Sherpa is working with a number of companies under the NDA in areas such as healthcare, and Uribe-Etekebaria said it plans to announce customers in other areas such as telecommunications, retail and insurance in the near future.