Quant jobs / Jane Street

Machine Learning Performance Engineer

A low-level performance role on the ML team, making model training and inference fast on GPU clusters. It covers both large-scale training and latency-sensitive inference in live systems.

What you’d work on

Our reading of the posting: what a new hire in this role would most likely spend their first months on.

Profile and debug training runs from storage through network to GPU
Write and tune CUDA kernels
Reduce latency for real-time inference
Improve cluster networking and collective communication setup

What they ask for

Required

Low-level systems programming and optimisation
Modern ML techniques and tools
GPU architecture knowledge
Profiling tools such as Nsight
GPU libraries such as cuDNN, CUTLASS and Triton
Distributed training with NCCL or MPI
English fluency

Preferred

High-speed networking such as InfiniBand and NVLink

Education: No degree level stated.
Experience: Experience in low-level systems programming and optimisation
Languages and tools: CUDA
C++
Triton
NCCL

How to prepare

Write a matrix-multiply kernel in CUDA and tune it with Nsight Compute
Study the GPU memory hierarchy and warp execution
Run a multi-GPU training job and measure communication overhead
Learn how collective operations like all-reduce work

What to study

The HFT Quant Developer roadmap sets out the stages for this kind of role, with free resources for each.

Open the roadmap

Apply

Read the full posting on Jane Street’s site before applying: requirements, location and pay can change after we check them.

View the posting on Jane Street’s site ↗

SelfMadeQuant wrote this summary from the employer’s public posting, last checked 6 November 2026. SelfMadeQuant isn’t affiliated with Jane Street and doesn’t handle applications.