Bulle Mostovoi

Bulle Mostovoi is a GPU DevTech engineer at NVIDIA, working on NVIDIA cuOpt and accelerated decision intelligence. She is the developer of mPDLP, a multi-GPU implementation of the Primal-Dual Hybrid Gradient method for solving large-scale linear programming problems using NVIDIA GPUs. Her work focuses on designing and optimizing GPU-accelerated optimization algorithms, with particular emphasis on multi-GPU parallelism, sparse matrix-vector multiplication, distributed computation, and performance scaling for large-scale linear programs. Through mPDLP and her work on cuOpt, she contributes to enabling optimization workloads that can leverage the computational capabilities of modern GPU architectures. Bulle holds an engineering degree from EPITA, where she specialized in image synthesis and GPU programming. Her technical interests include CUDA, high-performance computing, parallel algorithms, GPU architecture, sparse linear algebra, and mathematical optimization.