Too large to run locally, too sensitive for generic AI APIs. CloudFerro Sherlock delivers sovereign, secure inference, without building infrastructure from scratch.
CloudFerro Contribution in Ambiti8n project
CloudFerro contributes to Ambiti8n by providing sovereign cloud infrastructure and CloudFerro Sherlock, a secure Inference Platform for generative AI.
CloudFerro Sherlock is developed within the Gateway IPCEI-CIS project. It enables organisations to implement generative AI solutions in a structured, secure and operationally practical way, without building AI infrastructure from scratch. This helps teams move faster from concept to deployment.
In Ambiti8n, Sherlock provides a trusted inference path for large language models. The platform can support use cases such as knowledge assistants, document analysis, automated responses, advanced information search, coding support and language-based applications, including Polish-language solutions.
Sherlock models are available through an API that follows the OpenAI standard, making integration with existing environments and applications straightforward. Available models include Bielik v3.0, GPT-OSS 120B, PLLuM-8x7-chat, LLaMA 3.1 and 3.3, MiniMax M2.5 and Pixtral.
CloudFerro also supports the Ambiti8n demonstrator with cloud infrastructure used for hospital-side environments, integration and testing. This is important because not every AI model can or should run directly at the hospital edge. Larger models require more powerful infrastructure, while healthcare scenarios also demand strong data security, operational control and compliance with European regulations.
By combining infrastructure support with Sherlock’s sovereign inference capabilities, CloudFerro helps Ambiti8n demonstrate a practical model for generative AI in the European cloud-edge ecosystem. Sensitive data can remain under control, while applications gain access to advanced AI models through secure, easy-to-integrate APIs.
Turning Challenges into Opportunities
One of the important technical challenges in Ambiti8n was how to use advanced generative AI when the model is too large to run locally, but too sensitive to be handled through generic external AI APIs.
CloudFerro Sherlock addresses this challenge by providing a project-specific, sovereign inference environment. Large models can run on trusted CloudFerro cloud infrastructure and be consumed by the application through a secure, standard API. This avoids forcing all AI workloads into one deployment model and gives the architecture more flexibility.
For the application team, Sherlock turns complex model infrastructure into a practical service. They do not need to manage model hosting, scaling or the underlying cloud setup directly. They can focus on the clinical workflow and user experience, while Sherlock provides a reliable path to generative AI capabilities.
This approach is especially valuable in multi-partner and distributed architectures. Ambiti8n connects different components, providers and environments across borders. In such a setup, every service must be predictable, easy to integrate and clear in its operational role. Sherlock adds value by making sovereign inference available as a ready-to-use cloud service, rather than another infrastructure burden.
As a result, CloudFerro's contribution helps solve a concrete architectural problem: how to combine local execution, sovereign cloud inference and secure application integration in one working demonstrator.
How Collaboration Creates Added Value
For CloudFerro, Ambiti8n has been much more than a technical integration exercise. It has been a practical lesson in what it means to build within a European cloud-edge ecosystem.
Our role was to contribute Sherlock and the supporting cloud infrastructure, but from the beginning it was clear that these elements could not be developed in isolation. They had to become part of a wider chain: from data and model governance, through cataloguing and deployment, to the final clinical application. This required constant alignment with the other partners on interfaces, assumptions and the way each component would be used in practice.
The integration of Villanova into the CloudFerro Sherlock model catalogue is a concrete example of this value. It enabled a model developed by one of the consortium partners to become a lasting part of Sherlock’s resources, extending the benefits of the cooperation beyond the immediate project scope.
This reflects the values of IPCEI-CIS and 8ra: open collaboration, interoperability and European digital sovereignty. Ambiti8n demonstrates that advanced AI services can be built across different providers and countries, while remaining trusted, federated and practical to use.
For us, this is the lasting value of Ambiti8n: not only a working demonstrator, but relationships, shared experience and solutions that can continue to develop beyond the project itself.