ML Performance Engineer (HFT)

We are looking for an ML Performance Engineer to improve the speed and efficiency of large-scale model training in a high-performance trading environment. The role is focused on low-level optimization of ML training pipelines, GPU utilization, and distributed systems.

You will work close to the core infrastructure, where improvements in performance directly impact iteration speed and business results.

What you will do

• Optimize ML training workloads for speed, stability, and hardware efficiency.
• Profile and debug performance issues across GPUs, memory, and data pipelines.
• Apply CUDA, Triton, or similar low-level optimizations to improve training throughput.
• Improve data loading and preprocessing performance.
• Work on internal tooling for scalable and reproducible training.
• Collaborate closely with ML researchers and engineers on performance-critical systems.

What we are looking for

• Hands-on experience optimizing ML training or large-scale ML systems.
• Strong knowledge of PyTorch, JAX, or similar frameworks.
• Experience with deep learning models (e.g. Transformers, sequence models).
• Practical experience with CUDA / Triton / GPU profiling tools.
• Understanding of distributed training concepts and performance trade-offs.
• Strong Python skills.
• Comfortable working in fast-paced, high-impact environments.

Nice to have

• Experience with low-latency or performance-critical systems.
• Background in systems programming or HPC.
• Experience working close to hardware.

What we offer

• Competitive salary and performance-based bonuses.
• Modern hardware and technically challenging problems.
• High ownership and real impact on production systems.
• Flexible working style with minimal bureaucracy.
• Support for learning, conferences, and training.