Jul 3, 2025

Meet Dr Agnès Leroy, GPU Director at Cryptography Tool: Zama ZeeJuly 3, 2025 Zama is a cryptography company set out to redefine the standards for data privacy online. We are developing software tools to keep data private end-to-end, even during processing! This is made possible by a technology called Fully Homomorphic Encryption (FHE), a technique that enables data to be processed blindly without having to decrypt it. It is highly complex, but we are hiding all of its technical intricacies into easy to use tools, integrated with popular programming languages, so that any software developer can use it. I’m a software engineer and expert in Graphics Processing Unit (GPU) programming. My journey began as a researcher in Computational Fluid Dynamics at EDF (Electricity of France). I was working to improve a particle-based method to simulate water flows. Everything was running extremely slow, and I was determined to speed it up. I soon discovered GPUs were the key to solving these bottlenecks. After several years contributing to open-source projects, I realized my true passion was software development and made the switch. I then joined RTE (France’s Power Grid Operator), to work on a power grid simulation platform. Eventually, I heard about Zama and was intrigued by the challenge of improving Fully Homomorphic Encryption (FHE) performance with GPUs. Now, I lead a team at Zama, working to push the limits of FHE using GPU power. Zama is an incredible company in many ways: it is developing cutting edge technology that requires years of research and development. The potential is tremendous, and at the same time this is so new that the business around it is starting somewhat from scratch. I guess these are common challenges in Deep Tech, but getting to experience them first hand is incredible. Zama’s strategy to build the simplest possible tools to encourage adoption of the technology is key I think in this process. I can’t wait to see our tools used in mainstream applications: this will have deep implications on our digital and real lives I believe, and I’m very curious to see how it evolves. When I joined the company, ChatGPT had not been released yet. At the time we were mostly targeting AI applications, but the emergence of Large Language Models changed the game, as they require tremendous computational power. Introducing FHE to deliver new levels of data privacy in such models further increases this load, and at the moment the main challenge is to improve the computational speed. In the meanwhile, we started realising that Fully Homomorphic Encryption could bring data privacy in blockchain applications, which could have a huge effect on that sector. And our technology is already ready for that: that’s what we’ve been pushing for the past two years, with very exciting things to come in 2025. Zama is first set out to deeply change the way blockchains are operated, to safeguard users’ privacy. We believe this could be the missing piece to blockchain adoption in the conventional financial system, and we expect this to quickly gain pace. Beyond that, Zama’s ultimate goal is to redefine the https internet protocol into an httpz one, where every one’s data online is private by default. Related Articles