Machine Learning Performance Engineer

What you'll do

• Build scalable, robust training and inference pipelines for deep learning.
• Improve open-source deep-learning frameworks and identify and eliminate performance bottlenecks.
• Collaborate with researchers and engineers and develop an in-depth understanding of trading systems.

What they ask for

• Strong knowledge of low-level GPU programming with CUDA, including Tensor Cores, cooperative groups, graphs, and warp-level intrinsics.
• Expertise in deep-learning framework internals, such as PyTorch, JAX, or TensorFlow.
• Deep understanding of computer architecture.
• Experience in C++ and Python.

Nice to have

• Experience with the JAX ecosystem (XLA, Flax, etc.), GPU libraries and tools such as Triton, CUB, cuDNN, and cuBLAS, Linux system programming, or large-scale distributed training.
• Contributions to open-source projects related to data science and machine learning.

Pay

$200,000 — $200,000 USD (base salary)

Tools

CUDA, Tensor Cores, PyTorch, JAX, TensorFlow, C++, Python, Triton

AI focus

The role focuses on deep-learning training and inference infrastructure and performance for an AI initiative advancing futures trading strategies.