As the FHE Technology Director at Zama, I lead a focused team of researchers dedicated to pioneering the practical application of Fully Homomorphic Encryption (FHE) across various domains, including blockchain, machine learning, and more. Our mission is twofold: firstly, to pinpoint the specific requirements necessary for making FHE viable and efficient for diverse use cases, thereby unlocking its full potential for secure, privacy-preserving computations. Secondly, we are committed to staying at the forefront of cryptology research, diligently reviewing and assimilating the latest scholarly work into our development processes to ensure our solutions are cutting-edge. This involves not only integrating new findings from contemporary papers but also conducting our original research to address the unique challenges of FHE. Our ultimate goal is to bridge the gap between theoretical cryptology and practical, real-world applications, ensuring that FHE technology can be easily and effectively employed to protect data privacy in an increasingly digital world.

I contributed to the development of the Concrete project , which is an open-source Fully Homomorphic Encryption (FHE) compiler framework. This project enables developers to effortlessly create applications capable of performing computations on encrypted data without the need for decryption and prior knowledge. My contributions helped in simplifying complex cryptographic challenges, such as noise management and crypto parameter selection, thereby making FHE more accessible to developers and data scientists. I also contributed to the cryptographic library called TFHE-rs.

I delved into researching innovative neural network layers tailored specifically for the TFHE homomorphic encryption scheme. Our objective was to enhance secure computations, particularly when working with sensitive information. To test our solutions in practice, we participated in the iDash 2019 competition. This challenging event emphasized the use of neural network inference to process encrypted biological data, demonstrating the potential of our research in real-world applications. Throughout this project, I mostly used Python and C++.

I was tasked with enhancing the security of BPCE's mobile applications. I began by assessing the security measures of both existing and in-development apps. After identifying the key security features, I conducted a thorough risk analysis and determined the interconnections among these features. Subsequently, I developed proof-of-concept security solutions that met our standards, with a specific focus on TEE and Whitebox methodologies. This work was conducted using Python, Java, Objective-C, Kotlin, and Swift.

During a research project and an internship, I explored the statistical relationships between adjacent pixels to develop an integration strategy for steganography and created training sets minimizing the Cover Source Mismatch during tests on disjointed sets of images.

My primary objective was to use classifiers trained on large image datasets to identify photos altered using steganography. In this process, I examined how different parameters affected image creation. This work was conducted using Python and MATLAB.