Mehdi B.

Adaptation of a Zero-Knowledge Proof (ZKP) server for execution within AMD SEV-SNP Trusted Execution Environments, ensuring hardware isolation of cryptographic computations.

Extension of the system with GPU acceleration for critical ZKP primitives, including multi-scalar multiplication (MSM) and other large-scale parallel operations, under GPU confidential computing constraints.

Integration of NVIDIA GPU attestation to establish an end-to-end trust chain between CPU-side TEE execution and GPU-accelerated computations.

Mithril Security (acquired by H Company) (https://www.mithrilsecurity.io/) offers developers the ability to deploy "privacy-by-design" AI APIs.

Current contributions:

- Technical lead for the BlindAI project, enabling AI model deployment in Intel SGX enclaves. Recently received a positive security audit from the Quarkslab laboratory.

- Second technical lead on the BastionLab project, enabling secure data cross-referencing within enclaves.

- Technical lead for the BlindBox project, enabling AI API deployment in confidential VMs (AMD-SEV/AWS Nitro).

- Technical referent on projects carried out with clients (development of a "zero-trust" document search solution, deployment of AI models in enclaves...)

Realization of a POC to demonstrate Intel SGX capabilities for a potential integration into a blockchain project under development at Archethic. The POC included the implementation of attestation validation and a secret exchange protocol between enclaves, enabling secure communication and a unique network identity.