Machine Learning Performance Engineer

Overview As a Machine Learning Performance Engineer at Optiver, you shape scalable ML training and inference pipelines to support futures trading strategies. You will optimize performance across ML infrastructure, training, and inference, collaborating with researchers and engineers to reduce bottlenecks. You’ll dive into DL framework internals and low-level GPU programming to push efficiency at scale. This is a hands-on role in a fast-paced, research-driven environment where your work enables faster, more reliable trading insights. Compensation / Benefitshighly competitive compensation packageglobal profit-sharing pool and performance-based bonus401(k) match up to 50%comprehensive health, mental, dental, vision, disability, and life coverage25 paid vacation days plus market holidaysoffice perks: breakfast, lunch and snacks, social events, clubs, sporting leagues ResponsibilitiesBuild scalable training and inference pipelines for deep learningDive into internals of open-source deep learning frameworks and enhance their functionalityIdentify and eliminate performance bottlenecksCollaborate closely with researchers and other engineersDevelop an in-depth understanding of trading systems Key requirementsStrong knowledge of low-level GPU programming with CUDA, including Tensor Cores, cooperative groups, graphs, and warp-level intrinsicsExpertise in internals of deep-learning frameworks like PyTorch, JAX, TensorFlow, etc.Deep understanding of computer architectureExperience in C++ and PythonCollaborative mindsetproblem-solvingcontinuous improvement attitudeCUDA programmingPyTorchJAX