Imagine you want to study a secret recipe without ever seeing the ingredients list. This technology acts like a smart filter that lets researchers find patterns and train AI models while keeping the actual private data hidden. It creates a safe 'middleman' so companies can collaborate without risking a data leak.
Companies cannot use their most valuable sensitive data for AI training and research due to strict privacy laws and the risk of data leaks.
A privacy layer featuring a DP-LLM-FT module for fine-tuning LLMs, a DP-RAG module for queries, and an open-source SQL manipulation tool called Qrlew.
If you are a hospital network dealing with strict patient confidentiality laws — this project developed a privacy layer that allows researchers to analyze medical data without direct access. This removes privacy barriers while ensuring GDPR compliance.
If you are a bank dealing with sensitive financial transactions — this project developed a tool to fine-tune AI models using private data without exposing the training sets. This enables advanced analytics while protecting customer secrets.
If you are a city operator dealing with citizen movement data — this project developed synthetic data versions that facilitate urban analysis without compromising individual identities. This boosts trust in technology and societal progress.
Based on available project data, specific pricing is not listed, but the coordinator has already conducted successful paid pilots with hospitals and financial services companies.
Yes, the project's primary objective is to industrialize and robustify the solution to support any data workflow in a privacy-preserving way.
Based on available project data, some components like the DP-RAG module and the Qrlew tool have been open-sourced, while the core solution is being industrialized by Sarus Technologies.
The solution automates privacy protection by applying Differential Privacy principles, specifically designed to address concerns regarding GDPR and citizen rights.
The system is designed to blend effortlessly into existing workflows and includes Docker support for encapsulating computations in any language.
The project is led by a single SME, Sarus Technologies from France, which holds 100% of the industry ratio. This lean structure suggests a highly focused commercial drive, as the coordinator is directly responsible for both technical implementation and go-to-market initiatives.
Contact Sarus Technologies (FR) regarding their privacy-preserving data solution.
Contact us to explore how to integrate Differential Privacy into your AI pipeline.
