Sherpa.ai Federated Learning Platform

Sherpa.ai provides a federated learning platform and privacy-enhancing technologies for enterprises and governments. Its legal entity, Sherpa Europe S.L., is registered in Erandio, Bizkaia, Spain, according to the privacy policy. The founding year is not stated on the pages reviewed.

Models train locally inside each data silo, behind the customer's firewall. Encrypted model updates are aggregated into a global model, which can then be deployed at the edge. The platform supports machine learning, LLMs and multi-agent systems. Logos on the site include NIH, Telefonica, Prosegur, Zurich and Indra.

• Federated training: Models train locally in each data silo.
• Encrypted aggregation: Secure aggregation of model updates into a global model.
• Edge deployment: Deploy trained models for on-device inference.
• LLM and agent support: Covers machine learning, LLMs and multi-agent systems.
• SaaS platform: Collaborative AI model training delivered as a platform.
• Trust center: Security documentation published at trust.sherpa.ai.

Banks in a consortium can train a shared fraud model without exchanging transactions. Hospitals can build diagnostic models across sites. Insurers can improve risk models with partner data kept in place.

Data never leaves each organization's firewall, which eases privacy and sovereignty compliance. Enables joint model training across institutions that cannot pool data. Trained models can run at the edge for on-device inference. The site references ISO 27001 and HIPAA standards.

Federated learning is gaining attention as regulators and enterprises look for ways to train AI without centralizing sensitive data. Vietnamese banks, insurers and hospitals that hold siloed data and face data localization rules are natural candidates.

Sherpa Europe S.L., Erandio, Spain. 2026-07: raised 18 million USD with new investor Forgepoint Capital; existing investors Mundi Ventures, Ekarpen, Allegra Holdings and SETT. Recent contracts cited: Indra, NIH, Prosegur.