sherpa raises $8.5m to expand from conversational ai to b2b privacy-first federated learning services

Sherpa Pivots to Privacy-First AI for Enterprises
Based in Bilbao, Spain, Sherpa, a company initially focused on voice-based digital assistance and predictive search for Spanish speakers, has secured additional funding to concentrate on developing privacy-first AI services tailored for enterprise clients.
New Funding and Platform Development
The startup has completed an $8.5 million funding round. Founder and CEO Xabi Uribe-Etxebarria stated the capital will be allocated to expanding a machine learning platform prioritizing privacy, utilizing a federated learning model. This will complement their existing conversational AI and search capabilities.
Early adoption of the platform has been seen within Spanish public health services. They leveraged it to analyze COVID-19 case data, predicting emergency room demand and capacity nationwide.
Investors and Series Funding
This funding comes from Marcelo Gigliani, a managing partner at Apax Digital; Alex Cruz, chairman of British Airways; and Spanish investment firms Mundi Ventures and Ekarpen. It represents an extension to Sherpa’s previously raised $15 million in Series A funding.
Currently, Sherpa is also in the process of securing a larger Series B investment.
Shift in Focus: From Conversational AI to Federated Learning
The strategic shift towards building and commercializing federated learning services occurred as growth in the conversational AI sector began to plateau.
Early Success and Market Challenges
Sherpa initially gained traction with its Spanish-language voice assistant. This emerged during a period when major players like Apple (Siri) and Amazon (Alexa) hadn’t fully extended their services to non-English speaking markets.
By 2019, the service had surpassed 5 million users. Clients included prominent organizations such as the Spanish media company Prisa, Volkswagen, Porsche, and Samsung, utilizing its conversational AI and predictive search features.
Competitive Landscape and Strategic Realignment
As Uribe-Etxebarria explained, while the voice assistant continues to operate, a significant challenge arose. Larger companies eventually incorporated Spanish language support into their existing voice assistants.
The scale of investment from these major players made it unsustainable for Sherpa to compete effectively in the long term without a substantial partnership. “Competing against Amazon, Apple and others would have been impossible without a major deal,” he noted.
This realization prompted the company to explore alternative applications for its AI technology.
The Emergence of Federated Privacy
The exploration of federated privacy began when Sherpa considered expanding its predictive search services into productivity tools.
Uribe-Etxebarria described the challenge of enabling an assistant to process emails while maintaining user privacy. Federated learning was suggested as a potential solution to “teach” the assistant to work with email data. The initial team believed they could develop a system to read and respond to emails with a dedicated team.
Platform Success and Productization
The resulting platform exceeded expectations. After a year, Sherpa determined it could be adapted beyond email triage and offered as a product to others.
This would allow organizations to train machine learning models with sensitive data in a more privacy-compliant manner.
Competition in Federated Learning
Sherpa is not alone in this space. Google’s TensorFlow, Fate (supported by Tencent security experts), and PySyft, a federated learning open-source library, also utilize federated learning techniques.
Sherpa is currently collaborating with companies under non-disclosure agreements in the healthcare sector. Plans are in place to announce partnerships in telecoms, retail, and insurance in the coming months.
Early Stage Event Announcement
Early Stage is a leading event for startup entrepreneurs and investors. It provides practical guidance from successful founders and VCs on building businesses, securing funding, and portfolio management.
Sessions cover all aspects of company development, including fundraising, recruitment, sales, product-market fit, PR, marketing, and brand building. Each session includes dedicated time for audience questions and discussion.
Use code “TCARTICLE” at checkout for a 20% discount on tickets.
Ingrid Lunden
Ingrid's Professional Background
Ingrid served as a writer and editor for TechCrunch for over thirteen years, from February 2012 to May 2025. Her base of operations during this time was in London.
Early Career and Publications
Prior to her tenure at TechCrunch, Ingrid contributed to paidContent.org as a staff writer. She also maintained a consistent presence as a freelance writer for various publications.
Notably, Ingrid has authored articles for the Financial Times, demonstrating a broad range of journalistic experience.
Areas of Expertise
Ingrid’s reporting focuses on the dynamic fields of mobile technology, digital media, and advertising. She frequently explores the connections and overlaps between these industries.
Language Proficiency
While English is her primary language for professional communication, Ingrid possesses fluency in multiple languages.
- She is proficient in Russian.
- Spanish is another language she can utilize.
- French represents her third additional language skill.
Her language skills are ranked in terms of competence, with Russian being her strongest non-English language, followed by Spanish and then French.