Machine Learning Performance Engineer (London)

XTX Markets is a leading algorithmic trading firm using advanced machine learning technology to produce price forecasts for over 50,000 financial instruments across various asset classes. The firm trades over $250bn daily across 35 countries and employs over 250 staff worldwide.

We leverage talented professionals, modern computational techniques, and cutting-edge research infrastructure to analyze large data sets efficiently, enhancing our proprietary trading algorithms. Our models have evolved from econometrics to neural networks and deep learning architectures over the past decade.

XTX Markets possesses extensive computational resources, including over 25,000 GPUs and 650 petabytes of storage, with teams of world-class researchers from diverse backgrounds. We are building a large-scale data center in Finland to support our computational needs.

Our culture emphasizes teamwork, diversity, and excellence, with a non-hierarchical environment where everyone’s contributions are valued.

The Role

We seek an experienced engineer to support our Research Technology team. The role involves working at a low level to enhance the performance of machine learning training and inference, focusing on computation and networking acceleration. You will work on pushing hardware beyond manufacturer limits, developing a sophisticated optimizing compiler, and hand-optimizing real-world use cases.

Key responsibilities include creating optimal execution plans for complex computational graphs across various platforms, applying optimizations to intermediate representations, and improving backend code generation.

Attributes

You should be an experienced software engineer, proficient in C/C++/Rust or similar, with confidence in solving complex algorithmic problems. Additional desirable experience includes:

• Writing Linux kernel modules or OS development
• Targeting commodity GPUs or similar hardware
• Working with SSA-based IRs
• Understanding PCI Express architecture
• Developing hardware interface drivers