Andrei Stoian is a research scientist and ML engineer based in Paris with four years of experience specializing in privacy-preserving machine learning. At Zama he contributes to Concrete ML, enhancing FHE-compatible workflows by restoring benchmarks, advancing quantization-aware training for CNNs, and improving Brevitas integration and padding support. He blends hands-on engineering with applied research, shipping practical features that make encrypted inference more usable for real-world models. Notably, his work focuses on reconciling modern ML tooling with the constraints of fully homomorphic encryption, a niche that requires both systems thinking and deep model-level adjustments.
Concrete ML: Privacy Preserving ML framework using Fully Homomorphic Encryption (FHE), built on top of Concrete, with bindings to traditional ML frameworks.
Role in this project:
ML Engineer & Data Scientist
Contributions:529 reviews, 97 commits, 196 PRs in 1 year
Contributions summary:Andrei primarily focused on enhancing the `concrete-ml` library for privacy-preserving machine learning. They restored and improved benchmarks, specifically for quantized model evaluation. Their contributions included implementing and testing new features related to Quantization Aware Training (QAT) for convolutional neural networks, and debugging Brevitas integration. Additionally, the user improved existing neural network examples and introduced new functionality, such as support for padding, and fixed issues related to the underlying framework.
Concrete: TFHE Compiler that converts python programs into FHE equivalent
Contributions:44 reviews, 3 PRs, 17 pushes in 2 years 2 months
cryptanalysistfhefhezero-trustconcrete
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.