J
Posted 20 hours ago
Machine Learning Performance Engineer
Jane Street
Requirements
Modern ML techniques and toolsets, End-to-end training performance debugging, Low-level GPU knowledge: PTX, SASS, warps, Tensor Cores, memory hierarchy, CUDA GDB, NSight Systems, NSight Compute, Triton, CUTLASS, CUB, Thrust, cuDNN, cuBLAS, CUDA graph launch, tensor core arithmetic, warp sync, async memory loads, Infiniband, RoCE, GPUDirect, PXN, rail optimization, NVLink, NCCL or MPI collective algorithms, Inventive approach and willingness to ask hard questions
Skills
CUDAGPUTritonNCCLMPIInfiniBand
About the role
Responsibilities
- Optimize ML model performance for both training and inference
- Improve efficient large-scale training and low-latency real-time inference
- Take a whole-systems approach spanning storage, networking, host, and GPU levels
- Ensure platform efficiency down to the lowest level (caches, memory hierarchy)
Requirements
- Understanding of modern ML techniques and toolsets
- Systems knowledge to debug training run performance end to end
- Low-level GPU knowledge: PTX, SASS, warps, cooperative groups, Tensor Cores, memory hierarchy
- Debugging and optimization with CUDA GDB, NSight Systems, NSight Compute
- Library knowledge: Triton, CUTLASS, CUB, Thrust, cuDNN, cuBLAS
- Intuition about CUDA graph launch, tensor core arithmetic, warp sync, async memory loads
- Background in Infiniband, RoCE, GPUDirect, PXN, rail optimization, NVLink
- Understanding of collective algorithms in NCCL or MPI
- Inventive approach and willingness to ask hard questions
About the Company
Jane Street is a quantitative trading firm where machine learning is a critical pillar of the global business. The ever-evolving trading environment serves as a unique, rapid-feedback platform for ML experimentation.
ScoutJobs Agent
Get matches like this delivered daily
Sign up free — we'll pull jobs that fit your CV from across the web and rank them for you.
Get started — it's freeMachine Learning Performance Engineer
Jane Street · New York
