ML Performance Engineer: Real-Time GPU Systems Optimisation

Janestreet; Greater London; On-site; GBP 120,000 - 160,000; Full time

Job summary

Jane Street in London is seeking an engineer with a focus on low-level systems programming and optimisation for our ML team. You will optimize performance of models during training and inference across large-scale systems.

The role emphasizes real-time, low-latency inference, high-throughput research workloads, and a whole-systems approach from GPUs to storage and networking. Fluency in English and curiosity about finance are valued traits.

Qualifications

• Understanding of modern ML techniques and toolsets.
• Ability to debug a training run's performance end to end.
• Experience with low-level GPU memory hierarchy and PTX/SASS concepts.
• Experience debugging with CUDA GDB and Nsight tools.
• Familiarity with ML libraries such as Triton, CuDNN, CuBLAS.

Responsibilities

• Optimize model training and inference performance at scale.
• Improve real-time and high-throughput inference in large systems.
• Collaborate with ML and systems teams to deliver end-to-end optimization.

Skills

ML techniques

End-to-end perf debugging

Low-level GPU knowledge

CUDA debugging tools

cuDNN/cuBLAS familiarity

NCCL/MPI distributed training

Networking tech (Infiniband/NVLink)

English fluency

Tools

CUDA GDB

Nsight Systems

Nsight Compute

Triton

CUTLASS

CUB

Thrust

cuDNN

cuBLAS

Infiniband

RoCE

GPUDirect

NVLink

NCCL

MPI

Job description

Jane Street in London is seeking an engineer with a focus on low-level systems programming and optimisation for our ML team. You will optimize performance of models during training and inference across large-scale systems.

The role emphasizes real-time, low-latency inference, high-throughput research workloads, and a whole-systems approach from GPUs to storage and networking. Fluency in English and curiosity about finance are valued traits.